All references were classified and archived in Endnote X9. Data was extracted and recorded in a standard table using Excel 2013. Under the supervision and guidance of one author (JS), two other authors (YL, JY) simultaneously extracted the data from five randomly selected studies to assure quality in data extraction. The quality check of data extracted from all the included studies was performed by one author (JS) and confirmed by another author (COLU). After confirming that no data were missing, relevant data from all included studies were independently screened and extracted by two authors (YL, JY) for further analysis, including basic information about studies, methods, interventions, participants, outcomes, and overall findings, as listed in Table 2.

Two of the authors (YL, JY) independently assessed each included study using the 25-item version of the CONSORT-CHM 2017 statement (Cheng et al., 2017). Any disagreements in the assessment results were discussed or negotiated among them and later confirmed with the other two authors (JS, COLU). The CONSORT-CHM 2017 statement provided a grading system devised for each criterion that was used to determine the strengths and weaknesses of clinical trials of T&CMs interventions. According to the degree of conformity, the assessment results for each item were determined as non-existent, partially present (for example, if some aspects of the CONSORT project are missing or poorly described), and fully compliant.

Referring to the Cochrane guidelines, each study included was critically appraised independently by two of the authors (YL, JY) using the Cochrane Risk of Bias tool. The judgment was based on the definition of the recommendation by the Cochrane Handbook for Systematic Reviews of Interventions, and the assessment result for each item were grouped into one of the following three categories: “low risk of bias”, “unclear risk of bias”, and “high risk of bias”. Further explanation about the risk assessment is shown in Table 3.

Based on the PRISMA guidelines, screening process was conducted as shown in Figure 1. Initially, 6,170 records were yielded through searching of the seven electronic databases, and 33 records that met the criteria were found in the citations of other reviews, for a total of 6,203 records. After removing 4,098 duplicates, 3,985 articles were retained for further screening. During the process of screening by title and abstract, 2,191 records were excluded due to multiple reasons: completely irrelevant study or publication type (review or meta-analysis or protocol articles or animal experiments, etc.) (n = 649); focus on other diseases or symptoms of fatigue developed from other diseases: (n = 2,191); interventions using non-oral treatment other than T&CMs such as music, food, acupuncture, moxibustion or others (n = 1,061). Subsequently, the full text of 117 records were further assessed for eligibility of which 55 were excluded. Given the timeliness of the evidence, only trials published between 1 January 2013 and 31 December 2022 were included, resulting in 49 studies being excluded. To avoid a high placebo effect and to get the better validate effectiveness, single-arm RCTs were excluded (n = 6). Eventually, 62 trials were eligible for analysis in this review.

Among the 62 studies included in this review, 57 trials were published in Chinese and 5 trials were published in English. The RCTs were conducted in China (n = 58) (81-138) and Korea (n = 4) (Kim et al., 2013; Joung et al., 2019; Sung et al., 2020; Shin et al., 2021). All trials set a parallel design with double-arm or triple-arm. More details are shown inSupplementary Table S8. All but one trial (Dai et al., 2014) did not report specific information about the place or way of participant recruitment, including hospitals or medical centers. Among them, five studies were multi-center trials (n = 5) (Joung et al., 2019; Ma et al., 2019; Shin et al., 2021; Ma et al., 2022; Wu et al., 2022), and the rest were single-center trials (n = 56) (81, 84-98, 100-139, 141).

As shown in Supplementary Table S8; Table 4,195 participants, with 2,221 males and 2,842 females (the number of males and females included in 2 of the studies (Dai et al., 2014; Liu et al., 2021) were not detailed), were eventually allocated into the test groups (n = 2,704) and the control or comparison groups (n = 2,497). Participants in the included trials ranged in age from 15 to 75 years, with majority between 20 and 65 years. Four trials included adolescents with participants aged 15–44 years (Wang, 2014; Liu A. et al., 2015; Yang and Liang, 2016; Li, 2017). Three trials recruited middle-aged and older participants (45–70 years) (Li and Cao, 2015; Hu, 2019; Guo and Huang, 2022). Most included RCTs had no gender requirement. Two trials recruited female patients only (Zhao et al., 2013; Li and Cao, 2015) and two included male participants only (Sun et al., 2016; Wei et al., 2017).

Fifty-nine trials reported the history of participants with CFS, out of them, twenty-six trials only reported the range of participants’ history (Wang, 2014; Liu Y. et al., 2015; Liu A. et al., 2015; Gao and Pang, 2015; Li and Cao, 2015; Wang et al., 2015; Gao and Pang, 2016; Huang et al., 2016; Huang et al., 2017; Wei et al., 2017; Ou et al., 2018; Wu et al., 2018; Liu et al., 2019b; Liu et al., 2019c; Ding, 2019; Joung et al., 2019; Li et al., 2019; Lin et al., 2019; Ma et al., 2019; Kan et al., 2021; Liu et al., 2021; Shin et al., 2021). Twenty-two trials reported the range of participants’ history with mean history and standard deviation in the test and control groups (Xu D. et al., 2013; Xu Z. et al., 2013; Zhao et al., 2013; Liu et al., 2014; Guo and Guo, 2015; Li, 2015; Wu et al., 2016; Yang and Liang, 2016; Wang, 2017; Du, 2018; Li et al., 2018; Liu and Cai, 2018; Liu Y. et al., 2019; Hu, 2019; Mao, 2020; Sheng, 2020; Chen, 2021; Guo and Huang, 2022; Huang et al., 2022; Zhou et al., 2022). Three trials reported the average history of participants without providing the standard deviation (Su, 2013; Teng et al., 2014; Liu F. et al., 2019). Eight trials reported mean history and standard deviation in the test and control groups without the range of participants’ history (Kim et al., 2013; Niu et al., 2014; Wang, 2019; Dong, 2020; Sung et al., 2020; Li et al., 2022; Ma et al., 2022; Wu et al., 2022). The remaining three trials provided no information about the participants’ CFS condition (Li, 2017; Li, 2020; Zhang, 2020).

A total of 44 participants withdrew from ten trials (Zhao et al., 2013; Huang et al., 2016; Huang et al., 2017; Joung et al., 2019; Huang et al., 2020; Sung et al., 2020; Kan et al., 2021; Li et al., 2022; Ma et al., 2022; Wu et al., 2022), none of which exceeded 10% of the sample size in each trial, participant withdrew before intervention started in two trials, the rest dropped out during treatment, and only one trial reported specific reasons, details as shown in Supplementary Table S8, but did not report the reasons for dropout.

As shown in Table 4, among the included RCTs, the types of participants varied. Their eligibility was determined using different methods and criteria. Most of the participants recruited were diagnosed according the relevant standards of both Traditional Chinese Medicine and Modern Medicine (n = 41) (83-88, 91, 94-96, 98-102, 104, 105, 109-115, 118-125, 127-133, 135, 137), participants recruited for the included studies were only diagnosed for CFS according to the diagnosis standards in Modern Medicine (n = 15) (Kim et al., 2013; Su, 2013; Dai et al., 2014; Huang et al., 2017; Wu et al., 2018; Ding, 2019; Liu F. et al., 2019; Joung et al., 2019; Sung et al., 2020; Kan et al., 2021; Shin et al., 2021). In the remaining 6 studies, participants were recruited for diagnosis solely based on the relevant standards of Traditional Chinese Medicine (n = 6) (Li, 2017; Hu, 2019; Wang, 2019; Li, 2020; Mao, 2020; Zhou et al., 2022). Clear diagnostic details indicating CFS were presented in Table 4, including medical history enquiry, laboratory tests or physical examinations, as well as relevant fatigue scale scores to confirm the severity of CFS.

With Modern Medicine, the common diagnosis criteria reported in the studies was developed by the CDC in 1994 (Fukuda et al., 1994). Other tools from Modern Medicine used as diagnosis criteria included scales such as CIS score higher than 76 points at baseline (Shin et al., 2021), Korean version of the BDI score lower than 29 points and the Korean translation of the State–Trait Anxiety Inventory (STAI) score lower than 70 points (Kim et al., 2013), Hamilton Depression Scale (HAMD) scores on a scale of 8–35 point, and (or) Hamilton Anxiety Scale (HAMA) scores on a scale of 7–28 point (Huang et al., 2022). Laboratory tests (Dai et al., 2014; Liu et al., 2014; Wang, 2014; Liu A. et al., 2015; Yang and Liang, 2016) were used as another major diagnostic method in Modern Medicine, including complete blood count (CBC), urinalysis, erythrocyte sedimentation rate, electrolytes, blood glucose, liver function, renal function, thyroid-stimulating hormone. The diagnostic method applicable to both modern and traditional medical diagnostic was the investigation of past medical history (Dai et al., 2014; Liu et al., 2014; Wang, 2014; Liu A. et al., 2015; Yang and Liang, 2016; Yang et al., 2019; Dong, 2020; Guo and Huang, 2022), including clinical data inquiry. More details were shown in Table 4.

According to the theory of Traditional Chinese Medicine, fatigue in participants of the included studies was diagnosed based on the Guidelines for clinical research of new traditional Chinese medicines (2002) (Zheng, 2002), Clinical research guidelines for the treatment of chronic fatigue syndrome by new Chinese medicine drugs (2008) (Luo et al., 2008), State Administration of Traditional Chinese Medicine. Clinical diagnosis and treatment terminology and syndrome of traditional Chinese medicine (1997) (WHO, 1997), multiple versions of internal Medicine of Traditional Chinese Medicine textbooks (Zhou, 2007), and other literature or teaching materials. Referring to these criteria, participants were classified into the following 19 categories, namely, Liver depression and Spleen deficiency (n = 14) (Xu Z. et al., 2013; Gao and Pang, 2015; Guo and Guo, 2015; Li, 2015; Gao and Pang, 2016; Liu and Cai, 2018; Liu et al., 2019b; Liu et al., 2019c; Hu, 2019; Shi and Zha, 2019; Wang, 2019; Sheng, 2020; Chen, 2021; Liu et al., 2021), Qi deficiency (n = 6) (Teng et al., 2014; Wang et al., 2015; Li, 2017; Huang et al., 2020; Huang et al., 2022; Li et al., 2022), Heart and Spleen deficiency (n = 4) (Ou et al., 2018; Wu et al., 2018; Ding, 2019; Zhou et al., 2022), Qi and Blood deficiency (n = 3) (Zhao et al., 2013; Lin et al., 2019; Guo and Huang, 2022), Spleen and kidney deficiency (n = 3) (Wu et al., 2016; Wang, 2017; Liu Y. et al., 2019), Spleen and Qi deficiency, Liver depression and stagnation (n = 3) (Liu et al., 2014; Wang, 2014; Liu A. et al., 2015), Qi and Yin deficiency (n = 2) (Ma et al., 2022; Wu et al., 2022), Kidney deficiency (n = 2) (Li et al., 2019; Yang et al., 2019), Deficiency syndrome (n = 2) (Dai et al., 2014; Ma et al., 2019), Liver depression and Kidney deficiency (n = 2) (Niu et al., 2014; Sun et al., 2016), Spleen, Kidney and Yang deficiency (n = 1) (Mao, 2020), Spleen and Qi deficiency (n = 1) (Li, 2020), Spleen deficiency and humid heat (n = 1) (Dong, 2020), Kidney and Yin deficiency (n = 1) (Du, 2018), Liver depression, Spleen and Kidney deficiency (n = 1) (Zheng et al., 2017), Liver and Kidney deficiency (n = 1) (Wei et al., 2017), Yang deficiency (n = 1) (Yang and Liang, 2016), Liver depression and Blood deficiency (n = 1) (Liu Y. et al., 2015), Spleen and kidney deficiency, and Liver depression and stagnation(n = 1) (Li and Cao, 2015).

As shown in Supplementary Table S9, the combination of various ingredients was the main intervention investigated in the 62 included studies. Different forms of T&CM preparations were tested, including decoctions (n = 39) (82-86, 88, 90, 92, 93, 95, 97-100, 104, 106-111, 113, 115-119, 121-124, 126, 127, 129, 131, 132, 134, 136, 137), capsules (n = 5) (Kim et al., 2013; Sun et al., 2016; Joung et al., 2019; Sung et al., 2020; Zhou et al., 2022), plaster (n = 6) (Liu et al., 2014; Wang, 2014; Liu A. et al., 2015; Wei et al., 2017; Du, 2018; Huang et al., 2020), granules (n = 5) (Liu et al., 2019b; Liu et al., 2019c; Sheng, 2020; Liu et al., 2021; Shin et al., 2021), pills (n = 2) (Liu F. et al., 2019; Yang et al., 2019), tablets (n = 2) (Guo and Guo, 2015; Kan et al., 2021), drink (n = 1) (Li et al., 2019), mixtures (n = 1) (Li et al., 2019) and prepared slices (n = 1) (Su, 2013). Dosage forms including capsules, granules, pills, tablets, or mixtures of the T&CM tested group had clear quality standards or control. Most of the remaining T&CM interventions tested were formulated decoctions, and 16 studies mentioned that the basic medication could be adjusted according to the symptoms of individual participants during treatment and follow-up (Dai et al., 2014; Niu et al., 2014; Teng et al., 2014; Li, 2015; Wang et al., 2015; Wang, 2017; Du, 2018; Ou et al., 2018; Wu et al., 2018; Ding, 2019; Hu, 2019; Shi and Zha, 2019; Wang, 2019; Dong, 2020; Mao, 2020; Chen, 2021). Only eight RCTs of the included trials reported follow-up data, and four individual trials had follow-up longer than 6 months. Only one trial calculated recurrence rates without complete or detailed data.

The T&CMs investigated were mostly formulations of multiple herbal ingredients originated from Traditional Chinese Medicine. A total of forty-three formulas, including Buzhong Yiqi Decotion Combine with Xiaochaihu Decotion (n = 5), Chaihu Guizhi Decoction (n = 4), Wenzhen Yunqi Formula (n = 3), Guipi Decoction (n = 3), Jianpi Jieyu Xiaopi Paste (n = 3), Dispelling Dampness-Replenishing Qi-Nourishing Yin Step Therapy (n = 2), Modified Huangqi Jianzhong Decoction (n = 2). The following formulations were only used once in the included trials, namely, Xinshen’an Capsule, Guashen Decoction, Fali Decoction, Xiaoyao powder, Yishen Tiaodu Method, Yangwei Jianpi Plaster, Qingshu Yiqi Decotion, Zuogui Pills, Modified Xiaoyao San, Modified Erxian Decoction, Jianpi Yishen Decoction, Modified Lingzhi Pills, Self-made fatigue Decoction, Dalishen Tea, Bupiwei Xieyinhuo Shengyang Decoction, Yiqi Yangxue Bupi Hegan Decoction, Self-made Yishen Buxue Ointment, Shugan Jianpi Yishen Decoction, Long Gao, Bupi Yishen Decoction, Xiaopi - Yin, Shugan Yiyang Capsule, Wendan Decoction Combined with Sini powder, Yiqi Jianpi Bushen Decoction, Shugan Yangxue Method, Invigorating spleen warming kidney and smoothing liver Decoction, Danzhi Xiaoyao tablet, Shugan Jianpi Method, Buzhong Jiepi Decoction, Bushen Shugan Decoction, Buyi Pishen Decoction, Fufangteng Mixture, Modified Naoxinkang, Chaihu Combine with Longgu Muli Decoction, Panax quinquefolius L.prepared Slices. In the remaining 5 RCTs, Sipjeondaebo-tang (SJDBT), Korean red ginseng (KRG), Myelophil, Cistanche and Ginkgo extracts, and Extract of P. ginseng did not show any superior anti-fatigue effects when compared with placebo.

At least 148 ingredients were used in the included T&CM preparations for the treatment of CFS. Top 20 most common materials found in the T&CM preparations included Astragali Radix (黃芪), Glycyrrhizae Radix et Rhizoma (甘草), Atractylodis Macrocephalae Rhizoma (白朮), Bupleuri Radix (柴胡), Codonopsis pilosula (Franch.) Nannf. or C. pilosula Nannf.var.modesta (Nannf.L.T.Shen or Codonopsis tangshen Oliv) (党参), Paeoniae Radix Alba (白芍), Angelicae Sinensis Radix (當歸), Poria (茯苓), Citri Reticulatae Pericarpium (陳皮), Curcumae Radix (鬱金), Ziziphus jujuba Mill.(大枣), Pinellia ternata (Thunb.) Breit. (半夏), Rehmanniae Radix (地黄), Zingiber officinale Rosc. (生姜), Chuanxiong Rhizoma (川芎), Dioscoreae Rhizoma (山药), Ginseng Radix et Rhizoma (人參), Epimedii Folium (淫羊藿), Scutellaria baicalensis Georgi (黄芩)and Cinnamomum cassia Presl (桂枝).

As shown in Supplementary Table S8, among the included trials, the duration of the intervention ranged from 10 days to 4 months in the test and control groups, namely, 10 days (n = 1) (Xu D. et al., 2013), 2 weeks or 14 days (n = 2) (Wang, 2019; Zhang, 2020), 3 weeks (n = 1) (Shi and Zha, 2019), 4 weeks or 28 days or 1 month (n = 19) (Kim et al., 2013; Su, 2013; Gao and Pang, 2015; Li, 2015; Li and Cao, 2015; Gao and Pang, 2016; Wang, 2017; Liu and Cai, 2018; Liu et al., 2019b; Liu et al., 2019c; Liu F. et al., 2019; Li et al., 2019; Lin et al., 2019; Huang et al., 2020; Li, 2020; Sheng, 2020; Liu et al., 2021; Guo and Huang, 2022; Li et al., 2022), 6 weeks (n = 5) (Liu Y. et al., 2015; Wu et al., 2016; Du, 2018; Liu Y. et al., 2019; Sung et al., 2020), 8 weeks (n = 6) (Niu et al., 2014; Teng et al., 2014; Zheng et al., 2017; Mao, 2020; Shin et al., 2021; Zhou et al., 2022), 9 weeks or 60 days or 2 months (n = 7) (Yang and Liang, 2016; Li, 2017; Ma et al., 2019; Chen, 2021; Kan et al., 2021; Ma et al., 2022; Wu et al., 2022), 12 weeks or 90 days or 3 months (n = 12) (Zhao et al., 2013; Liu et al., 2014; Wang, 2014; Liu A. et al., 2015; Huang et al., 2016; Huang et al., 2017; Wei et al., 2017; Ou et al., 2018; Ding, 2019; Joung et al., 2019; Dong, 2020; Huang et al., 2022) and 4 months (n = 1) (Yang et al., 2019).The duration of the intervention described in three trials varied, namely, 15–30 days (n = 1) (Li et al., 2018), 4–8 weeks (n = 1) (Wang et al., 2015) and 30–40 days (n = 1) (Dai et al., 2014).

Two trials of the 62 studies had two control groups, therefore, the total number of control interventions was sixty-four, included placebo (n = 8) (Kim et al., 2013; Liu et al., 2019b; Liu et al., 2019c; Joung et al., 2019; Sung et al., 2020; Kan et al., 2021; Liu et al., 2021; Shin et al., 2021), pharmacological medicine (n = 54) or no treatment (n = 2) (Kim et al., 2013; Kan et al., 2021). For the 54 trials which tested T&CM against pharmacological medicine, conventional western medicine (n = 35) (Xu D. et al., 2013; Xu Z. et al., 2013; Su, 2013; Zhao et al., 2013; Niu et al., 2014; Teng et al., 2014; Liu Y. et al., 2015; Gao and Pang, 2015; Guo and Guo, 2015; Li, 2015; Li and Cao, 2015; Gao and Pang, 2016; Sun et al., 2016; Wu et al., 2016; Wang, 2017; Zheng et al., 2017; Li et al., 2018; Liu and Cai, 2018; Ou et al., 2018; Wu et al., 2018; Liu Y. et al., 2019; Liu F. et al., 2019; Hu, 2019; Li et al., 2019; Ma et al., 2019; Shi and Zha, 2019; Yang et al., 2019; Li, 2020; Mao, 2020; Guo and Huang, 2022; Huang et al., 2022) or traditional Chinese medicine patent prescription (n = 16) (Wang et al., 2015; Huang et al., 2016; Yang and Liang, 2016; Huang et al., 2017; Li, 2017; Wei et al., 2017; Lin et al., 2019; Wang, 2019; Dong, 2020; Huang et al., 2020; Sheng, 2020; Zhang, 2020; Li et al., 2022; Ma et al., 2022; Wu et al., 2022; Zhou et al., 2022), or traditional Chinese medicine combined with the conventional western medicine therapy (n = 3) (Liu et al., 2014; Wang, 2014; Liu A. et al., 2015) were used as comparators. Detailed information is provided in Supplementary Table S9.

As shown in Table 5, a total of 29 different scales were reported in sixty-two trials to measure the outcomes of CFS interventions. As the primary outcomes, FS-14 (n = 27) (Xu D. et al., 2013; Xu Z. et al., 2013; Liu et al., 2014; Niu et al., 2014; Liu Y. et al., 2015; Liu A. et al., 2015; Guo and Guo, 2015; Wang et al., 2015; Sun et al., 2016; Yang and Liang, 2016; Huang et al., 2017; Zheng et al., 2017; Du, 2018; Liu et al., 2019b; Liu et al., 2019c; Liu F. et al., 2019; Ma et al., 2019; Shi and Zha, 2019; Mao, 2020; Sheng, 2020; Liu et al., 2021; Guo and Huang, 2022; Huang et al., 2022; Li et al., 2022; Ma et al., 2022; Wu et al., 2022; Zhou et al., 2022) and TCM syndrome scores (n = 24) (Gao and Pang, 2015; Wang et al., 2015; Gao and Pang, 2016; Sun et al., 2016; Wu et al., 2016; Wei et al., 2017; Zheng et al., 2017; Du, 2018; Liu and Cai, 2018; Ou et al., 2018; Wu et al., 2018; Liu et al., 2019c; Ding, 2019; Hu, 2019; Li et al., 2019; Shi and Zha, 2019; Dong, 2020; Huang et al., 2020; Liu et al., 2021; Huang et al., 2022; Li et al., 2022; Ma et al., 2022; Wu et al., 2022; Zhou et al., 2022) were used in similar number of studies, and twelve of these trials used both scales simultaneously (Wang et al., 2015; Sun et al., 2016; Zheng et al., 2017; Du, 2018; Liu et al., 2019c; Shi and Zha, 2019; Liu et al., 2021; Huang et al., 2022; Li et al., 2022; Ma et al., 2022; Wu et al., 2022; Zhou et al., 2022). Secondary outcomes included the following categories of scales and fatigue-related measures. Different types of scales used were ranked from highest to lowest frequency of use.

Other fatigue scales: CFQ-11 (n = 5), FAI (n = 5), FSS (n = 4), MFI-20 (n = 3), FAS (n = 1), CIS (n = 1), ATMT (n = 1), as well as assists in measuring fatigue, VAS (n = 7), NRS (n = 2).; Measurements reflecting the status of fatigue or related functions: inflammatory factor (n = 9), biochemical indicators (n = 8), nuclear magnetic resonance (n = 2), and relapse rate (n = 1).; Other measurement scales: Mental health and mental condition scales: SAS (n = 6), SDS (n = 5), HAMD (n = 3), HAMA (n = 3), SCL-90 (n = 2), BDI (n = 1); Quality of life and health scales: SF-36 (n = 4), WHOQOL-BERF (n = 3), EQ-5D 5L (n = 2); Measurements of other items: PSQI (n = 4), QBYY-Q (n = 1), SLQ Questionnaire (n = 1), ISI (n = 1), Qi deficiency Constitution Scoring scale (n = 1), PSSG (n = 1), SRI (n = 1), ADL (n = 1).

Nineteen trials measured quantifiable physiological and biochemical indicators. Only nine reported statistically significant differences in pre- and post-intervention measurements between the test and control groups (Zheng et al., 2017; Du, 2018; Wu et al., 2018; Liu Y. et al., 2019; Liu et al., 2019c; Sheng, 2020; Li et al., 2022; Ma et al., 2022; Zhou et al., 2022), with the levels of serum immunoglobulin A (IgA), immunoglobulin M (IgM) and immunoglobulin E (IgE) reporting the most discrepancies.

A total of 58 out of 62 included trials reported overall effectiveness in their results, of which 53 trials calculated overall response rates. In the test groups, the overall efficacy rates were in the range between 77.7% and 98.0%, of which 37 trials reported more than 90% of overall efficacy rate (Xu D. et al., 2013; Xu Z. et al., 2013; Su, 2013; Liu et al., 2014; Teng et al., 2014; Wang, 2014; Gao and Pang, 2015; Guo and Guo, 2015; Li, 2015; Li and Cao, 2015; Gao and Pang, 2016; Sun et al., 2016; Wu et al., 2016; Yang and Liang, 2016; Li, 2017; Zheng et al., 2017; Du, 2018; Li et al., 2018; Liu and Cai, 2018; Wu et al., 2018; Liu Y. et al., 2019; Ding, 2019; Liu F. et al., 2019; Hu, 2019; Lin et al., 2019; Ma et al., 2019; Wang, 2019; Yang et al., 2019; Dong, 2020; Huang et al., 2020; Li, 2020; Mao, 2020; Sheng, 2020; Guo and Huang, 2022; Li et al., 2022; Ma et al., 2022; Wu et al., 2022) and only one trial with less than 50% overall efficacy rate (Shin et al., 2021). In the control groups, the overall efficacy rates ranged from 36.11% to 87.5%. The remaining six did not calculate the total efficacy rate. Instead, the difference in scale scores between the test group and the control group before and after treatment was used to assess the improvement in fatigue after interventions and showed statistical significance (p < 0.05) (Niu et al., 2014; Liu Y. et al., 2015; Liu A. et al., 2015; Huang et al., 2016; Huang et al., 2017; Zhang, 2020). Three trials did not report a statistically significant difference in fatigue improvement in the treatment group compared to the control group (Kim et al., 2013; Joung et al., 2019; Sung et al., 2020). Nevertheless, the trial of Sung et al. (2020) analyzing the subgroup with initial fatigue VAS below 80 mm and age greater than 50 years showed that the reduction in fatigue VAS was significantly greater if KRG was used instead of placebo (Sung et al., 2020). An analysis of a subpopulation of severe symptoms (NRS value ≥ 63), in the trial of Joung et al. (2019) showed a statistically significant improvement in fatigue symptoms in the Myelophil group compared with placebo (p < 0.05 for NRS, FSS, and SF-36) (Joung et al., 2019). In the trial of Kim et al. (2013), the NRS score of the ginseng dosing groups did not improve for physical fatigue symptoms, but significantly improved for mental fatigue symptoms. The ginseng groups also showed reduced ROS and MDA levels compared to placebo (Kim et al., 2013).

Only twenty-four out of 62 included studies monitored and reported any suspected adverse effects associated with the use of T&CM. Among them, 10 RCTs found no adverse reactions associated with the T&CM interventions tested (Kim et al., 2013; Huang et al., 2017; Ou et al., 2018; Liu et al., 2019b; Liu et al., 2019c; Li et al., 2019; Shi and Zha, 2019; Huang et al., 2020; Sung et al., 2020; Chen, 2021; Liu et al., 2021) while the remaining 14 trials had reported suspected adverse reactions. Among these 14 RCTs which reported adverse reactions, 4 RCTs only indicated the occurrence of adverse reactions and the types of adverse reactions (Joung et al., 2019; Guo and Huang, 2022; Ma et al., 2022; Wu et al., 2022), whereas the other 10 RCTs provided further information about the number of participants affected and whether they came from the test or control group (Xu Z. et al., 2013; Kim et al., 2013; Guo and Guo, 2015; Sun et al., 2016; Wu et al., 2016; Li et al., 2018; Liu and Cai, 2018; Kan et al., 2021; Shin et al., 2021; Zhou et al., 2022). All adverse reactions reported were non-serious and could mainly be divided into the following three categories: (Son, 2019): gastrointestinal discomfort (such as decreased appetite, diarrhea, stomach discomfort or mild pain, nausea, vomiting, diarrhea); (vant Leven et al., 2010); mild allergic reactions (dermatitis, systemic rash, itching); (Brurberg et al., 2014); and other minor complaints of cough, headache, blurred vision, dizziness, toothache, bleeding gums, flu-like symptoms, and other symptoms. Three trials reported discontinuation or disappearance of adverse effects after symptomatic treatment without affecting the results of the study.

The results of the quality assessment are summarized in Table 6. Only two of the 62 RCTs fully reported the 25 items required for the CONSORT-CHM. The two most common reasons for non-compliance were poor “Methods” (such as sample size, allocation, implementation, or blinding) and “Results” (such as participant flow, numbers analyzed, ancillary analysis or harms).

Only one trial did not report randomization (Dai et al., 2014), and of the additional 61 included trials, 22 trials mentioned randomization without specifying the method (Xu D. et al., 2013; Su, 2013; Dai et al., 2014; Liu et al., 2014; Niu et al., 2014; Teng et al., 2014; Wang, 2014; Liu A. et al., 2015; Li, 2015; Li and Cao, 2015; Yang and Liang, 2016; Huang et al., 2017; Li et al., 2018; Wu et al., 2018; Liu Y. et al., 2019; Liu F. et al., 2019; Ma et al., 2019; Wang, 2019; Yang et al., 2019; Sheng, 2020; Zhang, 2020; Li et al., 2022; Wu et al., 2022), and the remaining 39 trials reported the detailed randomization methods at low risk of bias, the most commonly used include, namely, random number table (n = 22) (Xu Z. et al., 2013; Zhao et al., 2013; Liu Y. et al., 2015; Gao and Pang, 2015; Guo and Guo, 2015; Wang et al., 2015; Sun et al., 2016; Wu et al., 2016; Li, 2017; Zheng et al., 2017; Li et al., 2018; Liu and Cai, 2018; Ou et al., 2018; Liu et al., 2019b; Li et al., 2019; Shi and Zha, 2019; Dong, 2020; Huang et al., 2020; Li, 2020; Guo and Huang, 2022; Huang et al., 2022; Zhou et al., 2022), and the block randomization (n = 10) (Kim et al., 2013; Gao and Pang, 2016; Liu et al., 2019c; Joung et al., 2019; Sung et al., 2020; Chen, 2021; Kan et al., 2021; Liu et al., 2021; Shin et al., 2021; Ma et al., 2022). Blinding was mentioned in seventeen studies, but only eight trials adequately described how this was conducted in detail (Liu et al., 2019b; Liu et al., 2019c; Joung et al., 2019; Sung et al., 2020; Liu et al., 2021; Shin et al., 2021; Ma et al., 2022; Wu et al., 2022). The remaining nine trials mentioned blinding but could not ensure the reliability of blinding (Kim et al., 2013; Wang et al., 2015; Sun et al., 2016; Wei et al., 2017; Liu Y. et al., 2019; Wang, 2019; Yang et al., 2019; Sheng, 2020; Kan et al., 2021). Forty-five trials mentioned no blinding procedures at all (81, 84-86, 88, 90, 92-95, 98, 99, 103-113, 115-119, 121, 122, 124-138). Only 6 studies provided information on allocation (Liu et al., 2019b; Liu et al., 2019c; Sung et al., 2020; Kan et al., 2021; Liu et al., 2021; Ma et al., 2022).

The overall risk of bias assessment results of all included trials is shown in Figure 2, and items of bias assessment for each included trial is shown in Figure 3. Most of the included studies provided incomplete information on study design and methodology. All of the included trials were found to be having unclear or high risk of bias in at least one domain. No trial was assessed as at low risk of bias in all domains. A detailed assessment of the seven domains at risk of bias was explained in the following: (Son, 2019): Randomization - only 34 trials were assessed as low risk of bias as all of them adequately reported the use of random sequence generation; (vant Leven et al., 2010); Allocation - only 8 trials reported detailed information on allocation concealment methods, which were considered sufficient to be assessed as low risk of bias, while 54 trials were assessed as unclear risk of bias for not providing specific method of allocation concealment or not describing them at all; (Brurberg et al., 2014); Blinding of participants and personnel - due to incomplete or easily interrupted blinding, 43 trials with blinding information were assessed as having high risk of bias. Three trials were assessed as having unclear risk of bias because no specific method of blinding of participants and personnel was provided; (Carruthers et al., 2011a); Blinding of outcome assessment -only 8 trials were assessed as low risk of bias as they reported detailed information on blinding of fact by outcome assessors, and blinding was not easily compromised; (Vercoulen et al., 1994); Incomplete outcome data - 38 trials reported little or no reporting of this information and were therefore assessed as trials at unclear risk of bias; (Matura et al., 2018); Selective reporting - 9 trials were assessed as having a high risk of reporting bias due to statistical errors or omissions in the results, such as incomplete or unclear interpretation of scale items; and (Jason et al., 1999) Other potential sources of bias - all 62 trials were assessed as unclear risk because of incomplete information reporting, such as failure to report sample size calculations or adverse effects, or lack sufficient justification or evidence to judge bias.

Among the 43 formulas investigated in the studies included in this review, the main intervention groups which showed significant benefits were Buzhong Yiqi Decoction combined with Xiao Chaihu Decoction (補中益氣湯合小柴胡湯) and Chaihu Guizhi Decoction (柴胡桂枝湯), which were significantly improved fatigue and the quality of life of participants. Regarding the treatment of fatigue with Buzhong Yiqi Decoction and Xiao Chaihu Decoction, the earliest trace could be traced back to the ancient Chinese medical book “Pujifang” in the Ming Dynasty (Chen et al., 2010). According to the Chinese traditional medical theory, most of the ingredients in the decoction were deficiency supplements, which were consistent with the systemic symptoms and signs of CFS which as liver depression and spleen deficiency, liver qi stasis, spleen deficiency, qi deficiency, blood stasis, liver and kidney yin deficiency, spleen and kidney yang deficiency (Luo et al., 2008).

In some included studies of the review, the T&CMs used were mainly formulations of multiple herbal ingredients originated from Traditional Chinese Medicine, with a wide variety and complex composition, including 148 ingredients. Moreover, in 16 studies, minor adjustments to herbal ingredients based on individual participants' fatigue symptoms (e.g., insomnia and depression) during the study period had been reported. It was called “Suizheng Jiajian” (隨證加減) or “Bianzheng Lunzhi” (辨證論治) with the theory of Chinese medicine (Chen et al., 2010). Compared to the fixed T&CMs preparations, information about the quality standards and the quality control of the T&CMs under investigation was often insufficient or ambiguous, if not lacking, which inevitably raised reasonable doubts about the safety of the T&CMs used by the participants. Reports of future RCTs should provide supporting adequate information that demonstrate the standardization of the T&CMs, such as composition, quality control, detailed dosing regimens and manufacturing processes.

Fatigue resistance can possibly be achieved mainly through reducing oxidant stress, regulating carbohydrate metabolism, delaying the accumulation of metabolites, promoting mitochondrial function, neuroprotection, anti-apoptosis, or regulating neurotransmitter disorder in the central nervous system. Most of herbal ingredients involved in this review included astragalus (Astragali Radix, 黃芪), angelica (Angelicae Sinensis Radix,當歸), paeony (Paeoniae Radix Alba,白芍), ginseng (Ginseng Radix et Rhizoma,人參), yam (Dioscoreae Rhizom,山藥), licorice (Glycyrrhizae Radix et Rhizom,甘草), all of which reportedly related to dopamine, hypothalamic-pituitary-thyroid axis disorders, compensatory effects generated by negative feedback inhibition, and activated immune-inflammatory pathways. For instance, increasing evidence has emerged in recent years that indicated the multiple immunomodulatory activities of Astragali Radix and Ginseng Radix resulting in therapeutic effects against fatigue in preclinical and clinical studies (Chen et al., 2010). Pharmacological studies also found that Astragali Radix could promote the recovery of fatigue by regulating glucose metabolism, lipid metabolism, and energy metabolism (Li et al., 2014). Ginseng Radix might inhibit oxidative stress and improve mitochondrial function in skeletal muscles (Bao et al., 2016). Bupleuri Radix had the functions of analgesic, antibacterial, antiviral, anti-inflammatory, anti-oxidation, and anti-depression. All of these mechanisms may have contributed to the positive impact on chronic fatigue (Jiang et al., 2020). Also, Paeonia Lactiflora was found to inhibit 5-HT synthesis and tryptophan hydroxylase expression and thus reduce fatigue during exercise and the resting state (Hong et al., 2003).

Methodological heterogeneity exists across all trials leading to questionable quality of the RCTs included in this study. Uncertainty about the risk of bias, inconsistent criteria for participant diagnosis and inclusion in RCTs, and inconsistent measures of T&CMs efficacy had made it impractical to perform meta-analysis for a comprehensive evaluation of the current evidence. For improvement in future studies, the following areas warrant careful consideration.

The use of the CONSORT-CHM statement was relatively consistent in only a few trials. The reporting quality of the rest of the studies was generally considered having unclear or high risk of bias. Most studies did not meet the recommended requirements, especially in the missing sample size calculation method, flow chart and other information, incomplete personnel allocation and blinding concealment information. Indeed, sample size calculations should be adequately performed and fully reported to ensure and demonstrate methodological quality. Otherwise, statistical bias, reduced trial quality, a lack of statistical capacity to correctly estimate treatment effects, and overestimation of the risk of intervention benefit were inevitable.

The evaluation results of the Cochrane risk-of-bias tool were also concerning in light of the multiple risks of bias associated with blinding. Although all trials reported that study protocols were designed according to recommended criteria, only eight RCTs reported comprehensive details of the blinding process, including participants (and personnel) and outcome assessment. Blinding is important to minimize bias and maximize the validity of study results (He et al., 2011). However, in other studies, researchers seemed to have failed to make improvements in study design and implementation, and the issue of blinding procedures for RCTs to test T&CM was repeatedly reported, as described in this study (Zhang et al., 2022).

According to the data analysis in this review, it was found that different inclusion and screening criteria of the participants also contributed to heterogeneity across the included studies. Over the past few decades, several countries and organizations developed new or improved diagnostic criteria (Jason et al., 2015). Importantly, most of these standard sets were primarily used for research purposes (e.g., epidemiology, pathophysiology, or therapeutic trials) rather than routine clinical practice (Jason et al., 2003; Sandler and Lloyd, 2020; National Institute for Health and Care Excellence, 2021). The most commonly used standard of Modern Medicine in this review were also drafted by an international expert group convened by the Centers for Disease Control and Prevention in 1994 (CDC) (Fukuda et al., 1994), which had been criticized for being overly accommodating to patients with milder disease (Reeves et al., 2003). Other guidelines and standards of traditional Chinese medicines were only used in China. However, these criteria had not been fully validated in extensive population studies.

In previous studies, the age, gender, and duration of CFS were representative factors affecting participants' function and symptoms (Chang et al., 2012; Stevelink et al., 2022). Women had about twice the risk of developing CFS than men (Faro et al., 2016), Moreover, patients who were older and with long-term CFS were more likely to experience reduced levels of vitality associated with physical fatigue, while short-term CFS patients were presented with worse levels of mental health (Kidd et al., 2016). Most of the included RCTs recruited participants in China, and only four RCTs were conducted in South Korea. These might impact on the generality of evidence and applicability of interventions to other populations.

A more comprehensive search strategy, stricter screening criteria should be set, ensuring the quality of the included studies, and focusing on other symptoms or causes to construct new diagnostic screening or criteria, and further evaluate the validity of the criteria to better build understanding of CFS disease. There is an urgent need to develop objective diagnostic tools for CFS. Insights into priority areas for improvement in conducting RCTs in T&CMs, such as standardizing the consensus of T&CMs in the diagnosis and treatment of CFS to establish unified comprehensive identification and efficacy measurement standards. It is expected that future RCTs could be improved by increasing awareness and adoption of the CONSORT-CHM guidelines. The CONSORT-CHM guidelines can be used to guide the design of the RCTs and as an assessment tool of the reporting quality. It is also important to conduct multicenter, large-sample RCT of CFS to eliminate influence caused by age and sex and to ensure the validity of results. Conducting clinical studies of multicomponent therapies to enrich the T&CMs studies of CFS and at the same time verify whether it has a comparative advantage over single component therapy in terms of safety and reliability may also be relevant to the practical needs in the clinical setting.

This review has a number of limitations. Firstly, this review only searched for RCTs within last 10 years, and did not further compare the results of different time periods. Therefore, it was not possible to assess the development of evidence about T&CMs for CFS over time. Secondly, due to the heterogeneity of results, it was not practical to conduct any meta-analysis to allow for more definitive conclusions. We also did not perform subgroup analyses to give more definitive answers, a shortcoming that should be addressed in our subsequent studies. Thirdly, the CONSORT-CHM extension was used in this review as an evaluation framework aiming to ensure the completeness and transparency in assessing the reporting quality of the RCTs included for analysis. However, few RCTs included in this review adopted the guideline. The overall reporting quality was found sub-optimal which might not necessarily reflect of the overall RCT design. Fourthly, the validated FS-14 and TCM syndrome score were commonly used to measure primary outcomes in the included studies. However, the subjectivity and ceiling effect of the tool itself further limit the reliability of the research results. The strength of the results of this review may be diminished due to these limitations, but this review remains relevant due to the growing popularity of studies of T&CMs interventions for CFS and the lack of conclusive evidence.