2.1 Characteristics of Included Studies.
We identified 332 potentially eligible articles. A total of 127 duplicates were excluded, and 150 unrelated records were also excluded by reading their abstracts. After reading 55 full-text articles, 38 records were excluded for reasons such as not a RCT, insufficiency of data and inconsistence of interventions or outcomes. Eventually, 17 RCTs[10-26] were included in the systematic review and 15 were included in the meta- analysis[10-11,13-25], all of which were published in Chinese. The study flow diagram is shown in Figure 1.
In this meta-analysis, a total of 2574 parturient women were selected, including 1288 cases in treatment groups and 1286 cases in control groups. The controls of all selected trials received obstetric care routine, such as psychological comfort and Doula delivery; the intervention measures in treatment groups included auricular acupressure (16 trials) and auricular acupuncture (1 trial). The main outcome measures were WHO pain level (5 trials) [15-16,19, 22-23], VAS score (4 trials)[10, 14,21, 24] and maternal satisfaction (3 trials)[10, 14,24]; the secondary outcome measures included duration of three stages of labour (1st 9 trials[10-11, 13-14,18-19, 21-24], 2nd 6 trials[13,16, 19, 22-23,25], 3rd 4 trials[13, 19,22-23]), postpartum blood loss (2 trials)[15, 25], neonatal Apgar score (4 trials)[10, 14,18, 24], number of spontaneous delivery cases (4 trials)[15-17,25] and number of caesarean section cases (3 trials)[15, 17,25]. Table 1 shows the main characteristics of the studies.
2.2 Methodological Quality of Included Studies.
Out of all the 17 randomized controlled trials, only 6 of them reported details about random sequence generation, 3 studies with a random number table tool and 3 studies with date. No studies mentioned any details related to allocation concealment. As for blinding, although 2 studies claimed they were single-blind trials, since the materials and manipulations used in the treatment were totally different in the test and control groups, it’s difficult to avoid performance bias. However, at least the outcome assessors should have been blinded, the detection bias of all studies was therefore classified as unclear. Most of the articles showed a low risk of incomplete outcome bias and selective reporting bias.