4-1. Results of the search
1974 possibly eligible studies were identified, and 12 additional studies were manually searched according to published reviews. After screening the title and the abstract, 1893 trials were excluded, and 93 papers were retrieved for full-text evaluation.
The authors were contacted of 5 studies, and 2 of them replied. Eventually, 22 articles meet the inclusion criteria for this review (Fig.1). The excluded 71 studies at full-text stage were listed in Table 7 with reasons.
4-2. Included studies
The characteristics of the included 22 studies are detaied in Table 6. Data from 4 studies were not available or incomplete in the content and no reply from the authors after requesting through e-mail(Lu and Chei, 2005, Flemmig et al., 1996, Jansson et al., 2003, Zingale et al., 2012). Of the 22 studies, 22 reported changes in PPD, and 20 reported changes in CAL. There were nine treatments among these 22 studies, and were listed as follows: Scaling and root planing or scaling alone (SRP/S, control), SRP/S + antimicrobial photodynamic therapy (aPDT), SRP/S + tetracycline fiber (TCF), SRP/S + metronidazole gel (MTZ), SRP/S + doxycycline gel (DOX), SRP/S + Minocycline gel (MIN), SRP/S + Minocycline microsphere (MIM), SRP/S + chlorhexidine chip (CHC), and SRP/S + chlorhexidine gel (CHG). Most of the studies did not record any severe
adverse events, only few studies reported minor side effects such as gingival swelling or gingival discomfort after administration of local antimicrobial agent(Radvar et al., 1996, Kasaj et al., 2007).
4-3. Risk of bias assessment
Risk of bias assessment of included studies was shown in Table 8. Since all the studies were included only if they were randomized controlled trials, the selection bias was assumed to be low. But unspecific information of random sequence generation and concealment of allocation were found to be unclear. Performance and detection biases were graded at high risk for open-labeled or unavailable double blinding trials. When the included study might have potentially conflicts of interest, it will be graded at high risk of other bias. As a result, most of the studies were at an unclear risk of bias.
4-4. Network meta-analysis
4-4-1 Changes of PPD in short-term follow-up
Network of eligible comparisons for changes of PPD in short-term follow-up group is presented in Fig. 2. Eight treatments yield in 28 pairwise comparisons, in which 10 of them were direct comparisons. In the short-term follow-up group, SRP + aPDT, SRP + TCF, and SRP + MTZ were the most frequently investigated treatments. Forest plot of treatment comparisons for PPD reduction is shown in Fig.8. Compared to SRP alone, SRP + TCF (0.56 mm, 95% CI: 0.29 ~ 0.82) and SRP + aPDT (0.45 mm, 95% CI: 0.06
~ 0.83) achieved significant reduction of PPD, whereas other antimicrobial agents only get insignificant additional PPD reduction. Table 3 shows summary of all pairwise comparisons by network meta-analysis in terms of PPD reduction and CAL gain in short-term follow-up. Compared to SRP alone, SRP in conjunction with CHC achieved the greatest additional PPD reduction than other antimicrobial agents (0.67 mm, 95%CI:
0.15-1.19). SRP + TCF and SRP + aPDT achieved 0.57mm (95%CI: 0.19-0.95) and 0.44mm (95%CI: 0.12-0.76) greater PPD reduction than SRP alone, respectively. SRP + MTZ, SRP + DOX, SRP + MIN, and SRP + CHG achieved additional 0.1 mm PPD reduction.
4-4-2 Changes of PPD in medium-term follow-up
Network of eligible comparisons for changes of PPD in medium-term follow-up group is presented in Fig. 3. Nine treatments yield in 36 pairwise comparisons, in which 13 of them were direct comparisons. In the medium-term follow-up group, SRP + aPDT, SRP + TCF, and SRP + MTZ were the most frequently investigated treatments. Forest plot of treatment comparisons for PPD reduction is shown in Fig.9. Compared to SRP alone, SRP + TCF (0.66 mm, 95% CI: 0.44 ~ 0.88) and SRP + aPDT (0.22 mm, 95%
CI: 0.01 ~ 0.42) achieved significant reduction of PPD, whereas other antimicrobial agents only get insignificant additional PPD reduction. Table 4 shows summary of all pairwise comparisons by network meta-analysis in terms of PPD reduction and CAL
gain in medium-term follow-up. Compared to SRP alone, SRP + CHC achieved the greatest additional PPD reduction than other treatments (0.65 mm, 95%CI: 0.21-1.10), SRP + TCF also achieved 0.64 mm (95%CI: 0.20-1.08), SRP + DOX and SRP + MIN achieved around 0.5 mm greater PPD reduction than SRP alone, while SRP+ MIM, SRP + MTZ, SRP + aPDT, and SRP + CHG achieved less than one third of a millimeter additional PPD reduction.
4-4-3 Changes of PPD in long-term follow-up
Network of eligible comparisons for changes of PPD in long-term follow-up group is presented in Fig. 4. Four treatments yield in 6 pairwise comparisons, in which 3 of them were direct comparisons. In the long-term follow-up group, SRP + aPDT was the most frequently investigated treatments. Forest plot of treatment comparisons for PPD reduction is shown in Fig.10. Compared to SRP alone, SRP + DOX (0.10 mm, 95% CI:
0.07 ~ 0.13) achieved significant reduction of PPD, whereas SRP + aPDT and SRP + MIM did not achieve additional PPD reduction. Table 5 shows summary of all pairwise comparisons by network meta-analysis in terms of PPD reduction and CAL gain in long-term follow-up. Only SRP + DOX achieved additional PPD reduction (0.10, 95%CI: -0.48-0.68) than SRP alone. SRP + aPDT and SRP + MIM achieved around 0.2 mm PPD increment than SRP alone.
4-4-4 Changes of CAL in short-term follow-up
Network of eligible comparisons for changes of CAL in short-term follow-up group is presented in Fig. 5. Eight treatments yield in 28 pairwise comparisons, in which 10 of them were direct comparisons. In the short-term follow-up group, SRP + aPDT, SRP + TCF, and SRP + MTZ were the most frequently investigated treatments. Forest plot of treatment comparisons for CAL gain is shown in Fig.11. Compared to SRP alone, SRP + aPDT (0.40 mm, 95% CI: 0.17 ~ 0.62), SRP + DOX (0.30 mm, 95% CI: 0.28 ~ 0.32), and SRP + TCF (0.28 mm, 95% CI: 0.10 ~ 0.45) achieved significant gain of CAL, whereas other antimicrobial agents only get insignificant additional CAL gain. Table 3 shows summary of all pairwise comparisons by network meta-analysis in terms of PPD reduction and CAL gain in short-term follow-up. Compared to SRP alone, SRP + CHC achieved the greatest additional CAL gain (0.48, 95%CI: 0.11-0.85), SRP + aPDT and SRP + TCF achieved 0.40mm (95%CI: 0.17-0.63) and 0.30mm (95%CI: 0.05-0.54) than SRP alone respectively. Other treatments achieved a slightly smaller additional CAL gain than SRP alone.
4-4-5 Changes of CAL in medium-term follow-up
Network of eligible comparisons for changes of CAL in medium-term follow-up group is presented in Fig. 6. Nine treatments yield in 36 pairwise comparisons, in which 13 of them were direct comparisons. In the medium-term follow-up group, SRP + aPDT, SRP + TCF, and SRP + MTZ were the most frequently investigated treatments. Forest
plot of treatment comparisons for CAL gain is shown in Fig.12. Compared to SRP alone, SRP + TCF (0.29 mm, 95% CI: 0.04 ~ 0.55) achieved significant gain of CAL. SRP + CHX achieved 0.77 mm additional CAL gain alone, but the confidence interval is between -0.23 and 1.77, thus did not reach statistical significance. Table 4 shows summary of all pairwise comparisons by network meta-analysis in terms of PPD reduction and CAL gain in medium-term follow-up. Compared to SRP alone, SRP + DOX achieved the greatest additional CAL gain (0.70, 95%CI: 0.09-1.31), SRP + CHC achieved 0.61mm (95%CI: 0.18-1.02) greater CAL gain than SRP alone, other
treatments achieved a slightly smaller additional CAL gain than SRP alone.
4-4-6 Changes of CAL in long-term follow-up
Network of eligible comparisons for changes of CAL in long-term follow-up group is presented in Fig. 7. Four treatments yield in 6 pairwise comparisons, in which 3 of them were direct comparisons. In the long-term follow-up group, SRP + aPDT was the most frequently investigated treatments. Forest plot of treatment comparisons for CAL gain is shown in Fig.13. Compared to SRP alone, SRP + DOX (0.20 mm, 95% CI: 0.15 ~ 0.25) achieved significant gain of CAL, whereas SRP + aPDT and SRP + MIM did not achieve additional CAL gain. Table 5 shows summary of all pairwise comparisons by network meta-analysis in terms of PPD reduction and CAL gain in long-term follow-up.
Only SRP + DOX achieved additional CAL gain (0.20, 95%CI: -0.43 ~ 0.83) than SRP
alone. SRP + aPDT and SRP + MIM achieved around 0.2 ~ 0.5 mm CAL loss than SRP alone.
4-5. Treatments ranking
We ranked the comparative effects of all antimicrobial agents with SUCRA (surface under the cumulative ranking) probabilities. For short-term PPD reduction, SRP + CHC ranked ahead (87.5%) for being the best intervention, SRP + TCF (83.4%) and SRP + aPDT (71.4%) had highest probability to be in the second and third ranking position respectively (Fig.14). SRP + DOX (30.6%) and SRP + MTZ (31.0%) were assigned to the last and second last position. For medium-term PPD reduction, SRP + CHC (79.4%) ranked for being the best intervention, SRP + TCF (78.4%) and SRP + DOX (67.5%) had highest probability to be in the second and third ranking position respectively (Fig.15). SRP + CHG (34.4%) and SRP + aPDT (35.9%) are the least desirable
treatment in PPD reduction. For long-term PPD reduction, SRP + DOX (68.3%) ranked as the best intervention, SRP + aPDT (40.3%) and SRP + MIM (33.7%) ranked as the second and third place (Fig.16).
For short-term CAL gain, SRP + CHC (85.6%) ranked ahead for being the best intervention, SRP + aPDT (78.1%) and SRP + TCF (64.2%) had the highest probability to be in the second and third ranking position(Fig.17). SRP + MIN (23.3%) and SRP + CHG (26.0%) were assigned to the last and second last position. For medium-term CAL
gain, SRP + DOX (85.9%) ranked ahead for being the best intervention, SRP + CHC (79.7%) and SRP + MTZ (51.9%) had the highest probability to be in the second and third ranking position(Fig.18). SRP + MIN (36.9%) and SRP + CHG (41.4%) are the least desirable treatment in CAL gain. For long-term CAL gain, SRP + DOX (81.2%) ranked as the best intervention, SRP + aPDT (40.8%) and SRP + MIM (16.0%) ranked as the second and third place (Fig.19).
4-6. Evaluating the inconsistency between direct and indirect evidence
In network meta-analysis, evidence from direct comparisons might be inconsistent with evidence from indirect comparisons, and validity of the meta-analysis may be affected. To evaluate inconsistency (design inconsistency, loop inconsistency, and side-splitting inconsistency) in different studies existed or not, we used a statistic model to check it. As a result, no statistical significant inconsistency in the analyses can be suggested among all the possible variables (Fig.20).
4-7. Comparisons between traditional pairwise and network meta-analysis
Table 1 compared the direct and indirect results in PPD reduction. Most of the point estimates displayed little differences, but the 95% confidence intervals of direct evidence and of network meta-analysis were overlapped substantially. The results of SRP + CHC compared to SRP alone revealed superior effect of SRP + CHC than SRP
alone both in pairwise and network meta-analysis. However, the confidence interval of pairwise meta-analysis has involved 0, which was not observed in network
meta-analysis. Table 2 compared results of CAL gain from pairwise meta-analysis to those of network meta-analysis. While there were small differences in the point estimates, the 95% confidence intervals were in general overlapped. Similar to results found in PPD reduction, SRP + CHC achieved greater effects than SRP alone in both kinds of meta-analysis, but the confidence interval of pairwise meta-analysis has involved 0 and this was not the case in network meta-anlysis.
4-8. Publicaiton bias
Publication bias might adversely affect the reliability of the conclusions of
meta-analysis. We therefore checked whether publication bias exists or not. Based on the funnel plot (Fig.21 ~ Fig.26), the graph is symmetric. Thus no evidence of
publication bias is suggested.