We searched seven major databases including PubMed, Embase, CINAH, the Cochrane Library, Wanfang, Chinese National Knowledge Infrastructure, and the Airiti Library from their inception to September 29, 2022, without language restriction. The search string used was based on MeSH and Emtree search headings: acupuncture (including acupuncture OR acupressure OR Shiatsu OR Zhi Ya OR Chih Ya OR Shiatzu OR auricular acupuncture OR ear acupuncture OR auricular acupressure OR ear acupressure OR auricular therapy OR auriculotherapy OR auricular needle OR otopoint OR otoneedle OR auriculoacupuncture OR otopuncture OR acupressure point OR acupoints OR Tui Na), pruritus, chronic kidney disease, dialysis, uremia. We also searched for free text words and word combinations containing the above mentioned terms (Supplementary Table S1). In addition to performing systematic searches of relevant databases, we conducted manual searches of the reference sections of the retrieved papers and contacted leading experts in the field to identify additional research studies. Furthermore, we examined unpublished studies available in the ClinicalTrials.gov registry¹ . This research is registered with PROSPERO (CRD42023425739). To ensure transparency and adherence to best practices, the reporting of this systematic review followed the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extended statement for network meta-analysis.

Randomized controlled trials (RCTs) were selected based on the following inclusion criteria: (1) inclusion of dialysis patients with UP, (2) utilization of APST, and (3) availability of quantitative data regarding itching severity. Articles were excluded based on the following criteria: (1) non-randomized study design, (2) absence of a UP diagnosis, and (3) inclusion of control groups receiving additional treatments such as oral Chinese herbal medicine, herbal baths, charcoal tablets, or antihistamines. In cases in which data were either missing or in raw form, we contacted the authors via email for clarification. In situations in which multiple articles presented overlapping data, we excluded duplicate articles and selected those with larger population sizes.

According to the above mentioned inclusion and exclusion criteria, two reviewers (Hui-En Chuo and Ping-Hsun Lu) evaluated the initially selected studies for eligibility for the network meta-analysis independently. The comments of the two reviewers were recorded and compared. Any disagreements were submitted and resolved by a third reviewer later. For each selected study, the following information was listed: first author, publication year, number of patients, patient ages, intervention modes of acupuncture, concomitant treatment, severity of pruritus, inspection data, effectiveness rate (ER), and acupoint.

Two reviewers independently performed a quality assessment of each study using the Risk of Bias 2 tool recommended by the Cochrane Collaboration (14). The assessment tool addresses five domains to evaluate the methodological quality of the included RCTs. Any difference of opinions between the reviewers was determined by a third reviewer.

Two outcomes were extracted and analyzed: visual analog scale (VAS) scores and ER of UP in dialysis patients. Higher VAS scores (0–10 points on each scale) indicate more severe pruritus. The definition of the ER is the number of patients with resolved pruritus or improvement in symptoms among the total patients which complain from UP.

To compare the groups, continuous consequences, such as VAS scores, were measured as weighted mean differences, and the ER was measured using dichotomous outcomes. We used the network package of STATA (version MP 17.0, StataCorp, College Station, TX, United States) to perform the statistical analyses as well as to generate the figures and tables (15). The surface under the cumulative ranking curve (SUCRA) is an index that ranges between 0 and 1; a larger value is associated with a better treatment response. Funnel plots were generated to detect publication bias. RevMan 5.4 (Cochrane Collaboration, Copenhagen, Denmark) was applied to evaluate the risk of bias graph and summary.

To evaluate the effect of APSTs for UP, RCTs were identified according to the PRISMA flowchart process (Figure 1). We identified 847 articles from electronic databases and excluded 632 articles based on the titles and abstracts. A total of 102 full-text articles were reviewed, and 75 of these articles were excluded for the following reasons: 15 studies were review articles, 17 studies involved different interventions, 18 studies were not RCTs, 2 studies did not involve patients with UP, 3 studies had abstracts only, 6 studies were protocols, 1 study provided incomplete data, and 13 studies did not report the target outcome. We synthesized the remaining 27 articles qualitatively and quantitatively.

Table 1 lists the characteristics of the retrieved RCTs. The included studies were published between 2004 and 2021 and had a total of 1,969 participants. The sample size of the included RCTs ranged from26–150 participants. All acupuncture interventions are combined with conventional therapy (dialysis for electrolyte balance, blood pressure, and body fluid maintenance). In studies following APSTs in patients with UP on dialysis, 11 involved acupuncture, nine involved AA, two evaluated patients were treated with AI and acupuncture (AI + A), and the remaining seven examined patients were treated with the following: acupoint far infrared (AFIR), AI, AI combined with AM (AI + AM), AIR, AM, AST, or ATENS. The acupoints used for each acupuncture therapy are listed in Supplementary Table S2.

The results of the risk of bias assessment charts for the included studies are shown in Supplementary Figures S1A,B. All the included trials were randomly allocated, and seven studies provided information about the hidden allocation sequence. Twelve studies described random methods (seven studies used tables of random numbers, three studies used random number generators, and one study used drawing of lots, and one study used chart numbers). Eight studies did not use intention-to-treat analysis, and the loss-to-follow-up rate of eight studies was greater than 5%. There were eight studies with lost outcome data >5%; none had any evidence to support the consistency of the missing data analysis results. Four studies reported different reasons for missing data, and the others did not report the reasons for missing data. Only two studies were double-blinded, and 1 study was single-blinded. The overall bias and outcome measurement domains exhibited a high risk of bias because most of the studies were not double-blinded and self-assessed itch scores are subjective outcomes. All outcome data were available from the included trials.

Sixteen studies including eight types of APST combinations in the network meta-analysis reported overall ERs. Six studies involved acupuncture, three involved AA, two involved AI + A, and the other involved the following different acupuncture treatment combinations: acupuncture with hemodiafiltration (A + HDF), acupuncture with hemoperfusion (A + HP), auricular acupressure with hemoperfusion (AA + HP), auricular acupressure with nocturnal dialysis 3 times per week (AA + ND), and AI + AM. For different APST combinations, the network graph (Figure 2) shows that the most publications evaluated traditional treatments, followed by APST, while A + HDF was the subject of few publications. Most trials reported comparisons between usual treatment (Tx) and acupuncture. Table 1 presents a comprehensive overview of the acupuncture treatment combinations that led to increased ERs through a direct comparison. In dialysis patients, the addition of different acupuncture treatment combinations to Tx, including Tx + A + HP, Tx + AA+ND, Tx + AI+AM, Tx + A, and Tx + AA, demonstrated a significant increase in the ERs of UP treatment compared with Tx alone. The treatment of Tx + A + HP (risk ratio (RR), 14.87; 95% confidence interval (CI), 2.18 to 101.53) was the most effective in increasing the ER, followed by Tx + AA+ND, Tx + AI+AM, and Tx + A, which were superior to Tx + AA (Table 1A). The overall ranking of each treatment can be visualized through the SUCRA. In Figure 3 and Table 2A, the descending order of SUCRA rankings for the ER of APST combinations in dialysis patients with UP are as follows: Tx + A + HP (SUCRA = 99.1), Tx + AA + ND (SUCRA = 86.3), Tx + AI + AM (SUCRA = 60.5), Tx + A (SUCRA = 54.2), Tx + A + HDF (SUCRA = 52.2), Tx + AA (SUCRA = 48.5), Tx + A + AI (SUCRA = 21.1), Tx + AA + HP (SUCRA = 20.3), and Tx (SUCRA = 7.8).

In the network meta-analysis, 16 studies reported VAS scores for 11 different combinations of APSTs. Among these, four studies used Tx + A, four used Tx + AA, and the remaining nine used various combinations, such as Tx + AFIR, Tx + AI, Tx + AI+AM, Tx + AIR, Tx + AST, Tx + AST + HPF, Tx + AST + HF, Tx + AM, and Tx + ATENS. In the network plot (Figure 4), it is evident that studies involving Tx + A and Tx + AA had the most subjects, whereas studies involving Tx + AFIR, Tx + AI, Tx + AI + AM, Tx + AIR, Tx + AST, Tx + AST + HPF, Tx + AST + HF, Tx + AM, and Tx + ATENS had only one subject per study. The majority of the trials reported a comparison between Tx and Tx + A. Table 1B, which shows the league table, presents the outcomes of both direct and indirect comparisons of the efficacy of various APSTs in decreasing VAS scores. The results indicate that Tx + A, Tx + AI + AM, and Tx + AA were significantly more effective in reducing VAS scores among dialysis patients with UP than Tx. The most successful intervention in reducing VAS scores among the tested treatments was Tx + A, with a mean difference (MD) of −2.63 (95% CI = −3.71 to −1.55). Following Tx + A, Tx + AI + AM was the second most effective treatment, with an MD of −2.04 (95% CI = −3.96 to −0.12). The ability of these treatments to reduce VAS scores of UP in dialysis patients was ranked based on SUCRA scores (Figure 5 and Table 2B). In descending order, the rankings were as follows: Tx + A (SUCRA = 88.0), Tx + AI + AM (SUCRA = 72.4), Tx + ATENS (SUCRA = 69.9), Tx + AIR (SUCRA = 66.9), Tx + AM (SUCRA = 63.1), Tx + AA (SUCRA = 61.8), Tx + AI (SUCRA = 47.4), Tx + AST + HF (SUCRA = 34.9), Tx + AST + HPF (SUCRA = 34.5), Tx + AFIR (SUCRA = 22.6), Tx + AST (SUCRA = 20.7), and Tx (SUCRA = 17.9).

The predicted SUCRA values for ERs and VAS scores were utilized to generate a cluster ranking plot (Supplementary Figure S2). Data on both ERs and VAS scores were available for the three APSTs: Tx + A, Tx + AI + AM, and Tx + AA. According to the analysis, Tx + AI + AM was expected to have a greater impact on improving both ER and VAS scores than Tx + AA. In addition, Tx + A was identified as the most effective treatment for improving the ER.

Supplementary Figure S3 displays the funnel plots that were generated to identify any publication bias in studies reporting ERs and VAS scores. The asymmetrical distributions observed in both funnel plots suggest the possibility of publication bias among the studies included in the analysis.