This meta-analysis was performed following the PRISMA 2020 guidelines for systematic reviews and meta-analyses.
Randomized controlled trials (RCTs) were included without restriction on language, publication type, or blinding.
Without restriction on age, race, and nationality. Patients in all age groups, with either HIV infection or diagnosed as AIDS patients and complicated with acute HZ will be the main inclusion criterion, with duration of HZ less than 3 weeks,
Trials testing any type of TCM therapies such as Chinese herbal medicine (CHM) or acupuncture, moxibustion, cupping with or without drugs (e.g. antiviral drugs, neurotrophic drugs, painkillers, and other common symptomatic treatments) were included. Combination of different type of TCM therapies as experimental treatment was also included.
The controls included no treatment, placebo and drugs therapy. If there were any drug therapy in control group, it must stay the same as intervention group.
If there were any co-interventions for HIV/AIDS, such as HAART or active antiretroviral therapy (ART), it should be same between control and intervention group.
1) Primary outcomes: Pain intensity (VAS score, McGill score, duration of pain); incidence of postherpetic neuralgia (PHN); cure rate (defined as a complete absence of pain and herpes) and cure rate defined as: number of cured/total number *100%);
2) Secondary outcomes: Quality of life (World Health Organization Quality Of Life - WHOQOL, Karnofsky Performance Status - KPS); adverse events.
Up to 23rd May 2021, we systematically searched the following electronic bibliographic databases: China National Knowledge Infrastructure (CNKI), Wanfang Database, Chong Qing VIP, SinoMed, PubMed, the Cochrane Library, WHO International Clinical Trial Registration Platform (WHO ICTRP) and ClinicalTrials.gov. The details of searching strategies are shown in Additional file 1.
All retrieved trials were managed using NoteExpress software. Initial screening was carried out based on inclusion/exclusion criteria after reading article titles and abstracts. The full-text screening was acquired and checked for eligibility before included in final analysis. Two authors carried out back-to-back literature screening independently. Any disagreement was solved through discussions and consultation with a third author.
The extracted data were: (1) basic characteristics of included trials (first author, research topic, year of publication, etc.); (2) diagnosis information; (3) inclusion and exclusion criteria; (4) baseline characteristics; (5) control and intervention details; (6) primary outcome and their definitions, measuring time points, follow-up, ; (7) risk of bias components.
The methodological quality of including trials were evaluated by the risk of bias assessment tool 2.0  developed by Cochrane Collaboration. This tool assess risk of bias from 5 domains: randomization process, intended interventions, missing data, outcome measurement, and selective reporting (valid/invalid). Only if there are 5 domains assessed as low, the overall risk as low, otherwise even 1 domain assessed as some concern/high, the overall risk will turn in some concern/high. And if there were 4 or more domains assessed as some concerns, the overall risk will turn in high.
Two authors (Y Jiang and RX Zheng) independently extracted data, assessed and verified the risk of bias. The results were cross-referenced, and any differences were resolved through discussion with the third author (J Li or JP Liu).
Review Manager 5.3 software  was used for data pooling and statistical analysis. According to the inclusion criteria of intervention and control, trials were divided into four comparisons. Meta-analysis was conducted for each comparison according to different outcomes. Mean difference (MD) and 95% confidence interval (CI) were used for continuous outcomes, and relative risk (RR) was calculated with 95%CI for dichotomous outcomes.
The Cochran Q test and the I2 statistics were used to examine the heterogeneity across the trials. The random effects model was used to estimate the size of the combined effects in the meta-analyses.
If P < 0.1 and the source of heterogeneity was unknown, descriptive analysis was used instead of meta-analysis. Subgroup analysis was used for moderate heterogeneity or higher (I2＞30% but less than 75%). Sensitivity analysis was used to investigate the effects of random-effects model analyses on heterogeneous results and the effects of any hypothesis.
Due to the large variability of TCM and the different combination of TCM and drugs, ineradicable clinical heterogeneity of the treatment measures in the included studies, the random-effects model was used to combine the data. The related charts for meta-analysis should be carried out, such as forest plot and funnel plot.