*3.8. Publication Bias*

A network funnel plot of the primary outcome (UPDRS-III) was constructed. There was no significant asymmetry seen in the visual inspection of the funnel plot (Figure 4). The Egger's test did not find any significant evidence of publication bias (*p* = 0.269). In the secondary outcome (UPDRS-II), there was no evidence of publication bias (Supplementary Materials Digital Content 7).

3.6.2. Comparative Effectiveness of the Acupuncture Modality in UPDRS-II

come, we did not conduct an additional sensitivity analysis.

*3.7. Adverse Events (AEs)* 

*3.8. Publication Bias* 

tary Materials Digital Content 7).

The probability of treatment as the best treatment option was presented through a measure called the P-score. The P-scores of the included modalities were as follows: BEEV (0.8971), ELEC (0.6685), MANU (0.5527), SHAM (0.3801), and CONV (0.0016). According to the P-score, BEEV was found to most likely be the best acupuncture modality for activities of daily life assessed by the UPDRS-II. The estimated effect size of each acupuncture modality compared to CONV via the NMA is presented in a treatment level forest plot and league table (Supplementary Materials Digital Content 7). In the treatment level forest plot and league table, the network estimate of the effect size (combining direct and indirect estimates) compared to CONV was as follows **(in favor of bold marks)**: **BEEV** (MD −6.07, 95% CI −9.41 to −2.72); **ELEC** (MD −4.50, 95% CI −6.19 to −2.80); **MANU** (MD −4.08, 95% CI −4.84 to −3.32); **SHAM** (MD −3.21, 95% CI −5.72 to −0.70). BEEV, MANU, ELEC, and SHAM were superior to CONV in the UPDRS-II. As UPDRS-II is a secondary out-

AEs were also assessed in the present study. Based on the comparisons, AE rates are summarized as follows. Reported AEs according to RCT design are as follows: 1) ELEC vs. CONV design: 3/30 AEs in the ELEC group vs. 12/30 AEs in the CONV group were reported (Chen 2012 [14]); 2) MANU vs. CONV: not reported; 3) BEEV vs. CONV: 1/18 AEs in the BEEV group were reported (Cho 2012 [41]); 4) ELEC vs. SHAM: not reported; 5) MANU vs. SHAM: 1/47 in the MANU group was reported (Kluger 2016 [43]); and 6) BEEV vs. SHAM: 4/20 AEs in the BEEV group were reported (Hartmann 2016 [39]).

A network funnel plot of the primary outcome (UPDRS-III) was constructed. There was no significant asymmetry seen in the visual inspection of the funnel plot (Figure 4). The Egger's test did not find any significant evidence of publication bias (*p* = 0.269). In the secondary outcome (UPDRS-II), there was no evidence of publication bias (Supplemen-

**Figure 4. Figure 4.** Network funnel plot: UPDRS-III score. Network funnel plot: UPDRS-III score.
