3.5.2. Comparative Effectiveness of the Acupuncture Modality in UPDRS-III

The probabilities of treatment ranking (P-score) among the included interventions were as follows: BEEV (0.9509), MANU (0.6325), ELEC (0.5349), SHAM (0.3685), and CONV (0.0132). According to the P-score, BEEV is most likely the best acupuncture modality for movement function assessed by the UPDRS-III (Figure 3 and Table 3). Mixed effect estimates (combining direct and indirect estimates) for each intervention compared with CONV were as follows **(in favor of bold marks)**: **BEEV** (MD −7.37, 95% CI −11.97 to −2.77); **MANU** (MD −4.13, 95% CI −5.78 to −2.47); **ELEC** (MD −3.66, 95% CI −6.29 to −1.03); SHAM (MD −2.71, 95% CI −5.92 to 0.50). BEEV, MANU, and ELEC were superior to CONV in UPDRS-III. However, SHAM was not statistically significant. No difference was observed in the comparison between the different acupuncture modalities (Table 3).

**Figure 2.** Network geometry of the included studies on UPDRS-III (Net graph). BEEV, bee venom acupuncture + conventional drug therapy; CONV, single conventional drug therapy; ELEC, electroacupuncture + conventional drug therapy; MANU, manual acupuncture + conventional drug therapy; SHAM, sham acupuncture + conventional drug therapy; UPDRS, Unified PD rating scale. **Figure 2.** Network geometry of the included studies on UPDRS-III (Net graph). BEEV, bee venom acupuncture + conventional drug therapy; CONV, single conventional drug therapy; ELEC, electroacupuncture + conventional drug therapy; MANU, manual acupuncture + conventional drug therapy; SHAM, sham acupuncture + conventional drug therapy; UPDRS, Unified PD rating scale. *Healthcare* **2021**, *9*, x 15 of 22

**Figure 3.** Treatment level network meta-analysis forest plot (UPDRS-III). BEEV, bee venom acupuncture + conventional drug therapy; CONV, single conventional drug therapy; ELEC, electroacupuncture + conventional drug therapy; MANU, manual acupuncture + conventional drug therapy; SHAM, sham acupuncture + conventional drug therapy; UPDRS, Unified PD rating scale. **Figure 3.** Treatment level network meta-analysis forest plot (UPDRS-III). BEEV, bee venom acupuncture + conventional drug therapy; CONV, single conventional drug therapy; ELEC, electroacupuncture + conventional drug therapy; MANU, manual acupuncture + conventional drug therapy; SHAM, sham acupuncture + conventional drug therapy; UPDRS, Unified PD rating scale.


**Table 3.** League table on UPDRS-III. **Table 3.** League table on UPDRS-III.

BEEV, bee venom acupuncture + conventional drug therapy; CONV, single conventional drug therapy; ELEC, electroacupuncture + conventional drug therapy; MANU, manual acupuncture + conventional drug therapy; SHAM, sham acupunc-BEEV, bee venom acupuncture + conventional drug therapy; CONV, single conventional drug therapy; ELEC, electroacupuncture + conventional drug therapy; MANU, manual acupuncture + conventional drug therapy; SHAM, sham acupuncture + conventional drug therapy; UPDRS, Unified PD rating scale.

ture + conventional drug therapy; UPDRS, Unified PD rating scale.

3.5.3. Sensitivity Analysis

sis (Supplementary Materials Digital Content 6).

3.6.1. Assumption of NMA and Network Geometry

*3.6. Secondary Outcome (Daily Life Activity, UPDRS-II): NMA* 

estimation of the direct and indirect comparisons of the effect size.

nificant results. The upper right triangle presents the effect size estimated using only direct comparison. As direct comparison does not exist in all treatment comparisons, there are several blanks in the upper right triangle. The lower left triangle provides a pooled

After excluding one study in the sensitivity analysis, 1) BEEV showed a tendency to be most effective in all 16 analyses; 2) in three sensitivity analyses (when excluding [40], [44], and [47]), the ranking between MANU and ELEC was changed, with ELEC showing a better effect; and 3) CONV tended to have the smallest effect size throughout the analy-

Homogeneity and transitivity assumptions are the same as those described in Section 3.5.1. We assessed the consistency assumption via a global and local approach and found no evidence of inconsistency after excluding the study by Lei [36]. The connectivity assumption was confirmed through network geometry (Supplementary Materials Digital Content 7). There were five nodes (ELEC, BEEV, MANU, CONV, and SHMA) from 10

studies and 14 pairwise comparisons from six types of comparison pairs (edges).

The part highlighted in BOLD with underlining is a comparison with statistically significant results. The upper right triangle presents the effect size estimated using only direct comparison. As direct comparison does not exist in all treatment comparisons, there are several blanks in the upper right triangle. The lower left triangle provides a pooled estimation of the direct and indirect comparisons of the effect size.
