Clinical and Preclinical Systematic Review of Panax ginseng C. A. Mey and Its Compounds for Fatigue

Background Fatigue, as a complex, multidimensional symptom, is associated with many physical illnesses. Panax ginseng C. A. Mey (PG) is an important herbal drug which has been used for benefiting Qi for thousand years. Panax ginseng C. A. Mey and its compounds (PGC) possess various pharmacological activities, including anti-fatigue. Here, we conducted a systematic review of both randomized clinical trials (RCTs) and preclinical animal studies to investigate the efficacy and safety of PGC for fatigue. Methods Electronic searches were performed in 7 databases from the time of each database's inception to August 2019. The methodological quality of RCTs was assessed using 7-item checklist recommended by Cochrane Collaboration or by the CAMARADES 10-item quality checklist. All the data were analyzed using Rev-Man 5.3 and Stata SE software. Results Eight eligible RCTs and 30 animal studies were identified. The risk of bias scores in RCTs ranged from 4/7 to 7/7, and of animal studies varied from 4/10 to 7/10. Meta-analyses showed that PGC was superior to placebo according to their respective fatigue scales, heart rate recovery, and clinical effect (P < 0.05). There were a similar number of adverse effects between PGC and placebo group (P > 0.05). Meta-analyses showed that PGC can significantly decrease level of blood lactate, blood urea nitrogen, creatine kinase, malondialdehyde, and lactic dehydrogenase in serum, level of malondialdehyde in liver and level of gamma-aminobutyric acid, 5-hydroxytryptamine in brain tissue, and increase swimming time, level of glutathione peroxidase, glucose, superoxide dismutase in serum, level of glycogen and activity of superoxide dismutase, glutathione peroxidase, and catalase in skeletal muscle, level of hepatic glycogen in liver and level of dopamine, acetylcholine in brain tissue, compared with control (P < 0.05). Meta-analyses showed no significant difference in animal body weight between PGC and control (P > 0.05). Conclusion The present findings supported, to a certain degree, that PGC can be recommended for routine use in fatigue. The possible mechanism of PGC resists fatigue, mainly through antioxidant stress, regulating carbohydrate metabolism, delaying the accumulation of metabolites, promoting mitochondrial function, neuroprotection, antiapoptosis, and regulating neurotransmitter disorder in central nervous system.


Description of the Condition
Fatigue is a condition of lacking the energy and motivation in responding to physical activity, emotional stress, boredom or insufficient sleep (Bach et al., 2016). It is a complex, multidimensional symptom that is prevalent in the general population (Jason et al., 2010). The cause of fatigue is unknown. The severity of fatigue varies greatly among individuals. Although fatigue does not lead to death, it has negative impacts on many areas of daily life (Arring et al., 2018). "Feeling weak all over for much of the time" was regarded as one of the most important symptoms (Lewis and Wessely, 1992). Most physical illnesses are associated with fatigue, such as many chronic diseases, namely anaemia, emphysema, asthma, and arthritis (Chen, 1986). Fatigue is a clinical challenge because its study of etiology, risk factors, and pathophysiology are still at an early stage. The goal of treatment is to treat symptoms and improve outcomes rather than to provide clear treatment (Alraek et al., 2011).

Description of the Intervention
Different interventions have been used in treating fatigue (Whiting et al., 2001). Nowadays, treatments commonly focus on muscle pain, sleep disorders, and emotional symptoms. Cognitive Behavior Therapy (CBT), various forms of exercise, as well as enhancement of coping ability, are standard treatment options. In addition, caregiver prescribed or self-administered medication are still common (Jones et al., 2007). It is indicated that no universal western medicine treatment can be recommended (Collatz et al., 2016). Interventions which include CBT and graded exercise therapy have shown promising results (Whiting et al., 2001). However, patients seem to be skeptical about CBT, who claimed that CBT and graded exercise for fatigue were neither effective nor safe (Twisk and Corsius, 2018). Recently, various forms of complementary and alternative medicine (CAM) have been widely used in fatigue such as herbal medicine, cheirapsis, balanced nutrition, and acupuncture (Jones et al., 2007). In particular, Panax ginseng C.A. Mey (PG) has been a rising utilization in treating fatigue in Asia and elsewhere around world. Based on traditional Chinese medicine and herbal philosophy, PG is considered as an adaptation to help restore body balance (Arring et al., 2018). Panax ginseng C.A. Mey is believed to improve overall quality of life (QoL), including energy and vitality, particularly during times of fatigue or stress (Yennurajalingam et al., 2017).

How the Intervention Might Work
Panax ginseng C.A. Mey has direct effects on the central nervous system (CNS), including cognition, sleep disorders, depression, pain, and the ability to regulate inflammatory cytokines (Yennurajalingam et al., 2017). To date, numerous active compounds have been identified such as ginsenosides, ginseng polysaccharides, and ginseng protein. Ginsenosides, the most important ingredients of ginseng, have been proved with various pharmacological activities such as anti-fatigue, anti-oxidation, neuroprotection, antiinflammation, and anti-diabetes. Ginsenoside Rg3 (Rg3) is one of the most abundant ginsenosides. It may improve exercise performance and increase fatigue resistance by enhancing deacetylase activity of silent information regulator of transcription 1 (SIRT1) and inhibiting the transcriptional activity of p53 (Yang et al., 2018). Ginseng polysaccharides have anti-fatigue activity probably by mobilizing triglyceride (TG) or fat during exercise, or by changing the activities oflactic dehydrogenase (LDH), malondialdehyde (MDA) and glutathione peroxidase (GPH-Px) to avoid lipid oxidation and protect corpuscular membrane . Ginseng proteins could resist fatigue through retarding the accumulation of blood lactate (BLA) and blood urea nitrogen (BUN), enhancing hepatic glycogen levels, and improving the ability of antioxidant enzymes .

Why It Is Important to Do This Review
Panax ginseng C.A. Mey is one of the most widely used plant products worldwide which has been used in oriental countries for thousands of years . Based on a comprehensive collection of clinical trials, systematic review can perform comprehensive analysis and statistical processing on qualified studies to form relatively reliable results, which can guide clinical decision-making. In addition, systematic review can solve the following clinical problems: research on the effectiveness of treatment, evaluation of diagnostic methods, prognosis estimation, analysis of the cost and benefit of treatment. Up to now, at least 2 systematic reviews have been conducted to evaluate efficacy and safety of Panax ginseng C. A. Mey and its compounds (PGC) for fatigue (Bach et al., 2016;Arring et al., 2018). However, the results of these reviews are inconclusive because of methodological flaws in their included primary studies. Cochrane group have developed an extensive set of guide lines for systematic reviews. These "not-so-good" studies were excluded with a strict process (Xie et al., 2013). In addition, the efficacy and mechanisms of PGC in fatigue animal models have not been systematically evaluated yet. Systematic review of animal researches is indispensable in the process of drug development and elucidation of the physiological and pathological mechanisms (Zheng et al., 2018). Preclinical research is the key to convert preclinical data into clinical data. In addition, systematic review of animal research is a more economic and ethical approach, which can integrate preclinical evidence, help reduce unnecessary sacrifice of laboratory animals, and prevent ineffective or less informative research (Zhou et al., 2019). As we all know, there is a gap between clinical research and clinical practice. More communication is needed between animal researchers and clinical researchers. Systematic review of animal experiments can lead to better collaboration between research groups and encourage the use of iterative methods to improve the relevance of animal models to clinical trial design. If the model cannot well simulate the clinical situation, it can be adjusted accordingly. In addition, as in human research, systematic review helps to identify and improve behavioral and reporting deficiencies in animal research Perel et al., 2007. Systematic review can effectively integrate preclinical comprehensive evidence and guide potential clinical translation. Thus, the aim of present study was to systematically summarize and critically evaluate the data from randomized control trials (RCTs) and animal studies of PGC for fatigue.

Search Strategy
This study followed the PRISMA statement (Stewart et al., 2015). EMBASE, PubMed, Cochrane Library, China National Knowledge Infrastructure (CNKI), VIP database (VIP), China Biology Medicine Database (CBM) and Wangfang database were electronically searched from their inception to August 2019. The following keywords were used: "fatigue OR Lassitude lethargy OR exhaustion OR weariness OR tiredness" and "panax OR Ginseng OR renshen" in Chinese or in English. All searches were limited to animal studies and clinical trials.

Eligibility Criteria
The prespecified inclusion criteria of RCTs listed below: (1) RCTs that evaluated the effectiveness and safety of PGC for fatigue; (2) the Cochrane risk of bias (ROB) tool met at least 4 out of the 7 domains; (3) Subjects had chronic fatigue syndrome (CFS) or healthy adults after exercise; Subjects were classified as CFS-like according to Evaluation and Classification of Unexplained Chronic Fatigue (ECUCF) (Fukuda et al., 1994); (4) PGC as monotherapy was used as an intervention in the treatment group, and interventions for control group were placebo or vehicle treatment; (5) The primary outcome measures were scales of fatigue and/ or objective evaluation criteria (e.g. physical performance, biochemical parameters). The secondary outcome measures were clinical effect according to fatigue scales and adverse events. The exclusion criteria were prespecified as follows: (1) fatigue caused by a medical condition, or withdrawal from medicines or substance; (2) duplicate publications and no available data.
The inclusion criteria of animal studies were prespecified as follows: (1) PGC for fatigue animal models was established by forced movement; (2) The interventions of treatment group were PGC at any dose and control group were nonfunctional liquid (normal saline) or no treatment; and (3) The primary outcome measures were forced movement time and/or serum biochemical parameters and/or skeletal muscle biochemical parameters and/or liver biochemical parameters and/or brain tissue biochemical parameters. The secondary outcome measures were body weight, organ index (organ weight/body weight) and possible mechanisms of PGC for anti-fatigue. The exclusion criteria were predefined as follows: (1) not fatigue model; (2) combined use of other drugs; and (3) no available data, duplicate publications, and lack of control group.

Data Extraction
Two independent researchers extracted the details from the included RCTs and animal studies according to two standardized data extraction forms, respectively. There are many manners for including outcomes, such as peak time point, last time point, and same time point. There is undeniable that any manner will lead to bias. In order to minimize bias, inclusion criteria were prespecified as follows: The result of the peak time point was included when the data were expressed at different times. If meta-analysis data were lost or expressed graphically, we would try to contact the author for more information. When no response was received, we used digital ruler software or exclusion software to measure data from charts. If the data in the primary RCT were missing or merely illustrated graphically, an effort was launched to obtain further information through contacting the authors. If failed, the digital ruler software was used for measuring data from the graphs or excluded.

Quality Assessment
The methodological quality of the included RCTs was evaluated by using the Cochrane Collaboration's tool. The RoB of the included animal studies was assessed using 10-item quality checklist of the Collaborative Approach to Meta-Analysis and Review of Animal Data from Experimental Studies (CAMARADES) with minor modification. Divergences were well settled by correspondence author (GZ).

Statistical Analysis
Analysis was conducted with RevMan 5.3 and Stata SE software. Continuous outcomes were presented as mean difference (MD) or standardized mean difference (SMD) with 95% confidence interval (CI). Dichotomous outcomes were presented risk ratio (RR) or odds ratio (OR) with 95% CI. Probability values P <0.05 were considered significant. In order to estimate heterogeneity across studies, we used I 2 -statistic test. An I 2 value greater than 50% was considered as having substantial heterogeneity. When substantial heterogeneity was not observed, the fixed-effects model was reported. On the contrary, the random-effects model was reported. Simultaneously, considering the differences in subjects, interventions, and treatments, we used the Z-test for subgroup analysis. P <0.05 was considered to be statistically significant. If an outcome contained more than 10 RCTs, funnel plots, and Egger's test were used to examine publication bias.

Study Selection
A total of 1331 relevant literatures were retrieved from the database, of which 667 were considered duplicates. Of the remaining 664 articles, qualified RCTs and animal experiments should be selected separately. For RCTs, 362 articles were eliminated because of reviews, case report, or animal studies. After scanning the remaining 302 full-text articles, 294 studies were excluded by reasons that they were (1) combined with other disease; (2) combined with other herbal treatment(s) in the intervention group; (3) no data available; (4) not real RCTs or quasi-RCTs; or (5) with the less than 4 domains "yes" according to the Cochrane RoB tool. Eventually, 8 RCTs were selected ( Figure 1A). For animal studies, 243 studies including clinical trials, case reports or review articles were excluded. Through fulltext evaluation of the remaining 421 studies, 391 were excluded for at least one of the following reasons: (1) lack of control group; (2) inappropriate fatigue model; (3) combined with other herbal treatment(s) in the intervention group; (4) unavailable data. Ultimately, 30 animal studies were included ( Figure 1B).

Study Quality
Randomized Clinical Trials Table 5 illustrates the methodological quality of 8 RCTs based on the Cochrane Collaboration's tool. All of them were ranged from 4 to 7 points. All included studies reported the method of random sequences generation, the criteria of a double-blind study design, and taking the complete outcome data into account. Three studies (Hartz et al., 2004;Hyeong-Geug et al., 2013;Lee et al., 2016) reported using allocation concealment. Two studies (Hartz et al., 2004;Lee et al., 2016) applied blinding specifically during outcome measure assessment. The protocols of 3 studies (Hyeong-Geug et al., 2013;Kim et al., 2016;Lee et al., 2016) were registered in the Clinical Trial Registry. In other bias, all eight studies were supported by nonprofit institutions and accounted for baseline comparability, but no study provided sample size estimation information.

Publication Bias
Funnel plots were conducted for two outcomes ( Figures 9B, C).
The results showed symmetrical distribution for the outcomes of BUN levels (Egger's test t = −1.05; 95% CI, −6.93 to 2.50; P = 0.320), which did not suggest an obvious publication bias. However, there was a significant bias in the BLA levels with Egger's test (t = −3.47; 95% CI, −5.14 to −1.19; P =0.004). Because the number of studies in the remaining outcomes was limited (n < 10), funnel plot and Egger's test were not appropriate.

Possible Mechanisms
PGC improved activity of GSH-Px, CAT, and SOD, scavenged free radicals and their metabolites, reduced the excessive ROS, and decreased levels of MDA, CK, and LDH. PGC decreased nitric oxide synthase (NOS), reduced toxic oxidant peroxynitrite, and prevented mitochondrial dysfunction and lipid peroxidation. PGC may enhance fat mobilization and promote gluconeogenesis, increase the delivery of glucose, and maintain blood glucose level. PGC increased the LDH activity and the hepatic glycogen levels, and retarded the accumulation of BUN and BLA. PGC improved succinate dehydrogenase (SDH), Na + -K + -ATPase, and Ca 2+ Mg 2+ -ATPase activities, enhanced mitochondrial function, and produced more adenosine triphosphate (ATP). PGC attenuated MPP + -induced MPTP-induced and apoptosis. PGC increased Ach and DA levels, and decreased GABA and 5-HT levels in brain tissue ( Figure 10).

Summary of Evidence
This is the clinical and preclinical systematic review to evaluate the efficacy and safety of PGC for fatigue. Eight RCTs with 678 participants and 30 studies with 2249 animals were selected. The quality of RCTs included was high, and animal studies were generally moderate. The findings of RCTs demonstrated that PGC was superior to placebo according to their respective fatigue scales, heart rate recovery, and clinical effect. There were a similar number of adverse effects between PGC and placebo group. The evidence available from animal studies showed that PGC could preserve physical function after exercise, mainly through antioxidant, regulating carbohydrate metabolism, delaying the accumulation of metabolites, promoting mitochondrial function, neuroprotection, antiapoptosis, and regulating neurotransmitter disorder in central nervous system.

Limitations
Some methodological flaws existed in the primary RCTs. First of all, 3 studies used allocation concealment (Hartz et al., 2004;Hyeong-Geug et al., 2013;Lee et al., 2016). Trials with unreported or inadequate allocation concealment could be exaggerated an average 18% beneficial effect of interventions (Higgins and Green, 2011). Second, non-blinding of outcome assessment may lead to systemic errors. In previous reviews, nonblinding of outcome assessment can lead to an overestimation of treatment effect by 27% to 68%, depending on different outcome types, i.e., binary outcome, measurement scale outcome, and time-to-event outcome. However, in RCTs, blinded outcome assessment was commonly poor reported and used. In present study, only 2 RCTs (Hartz et al., 2004;Lee et al., 2016) reported blinding of outcome assessment. The observer bias can be avoided by sufficient blinding. More independent assessors can be further used to increase the feasibility of blind assessment (Brennan et al., 2015). Thirdly, 3 RCTs (Hyeong-Geug et al., 2013;Kim et al., 2016;Lee et al., 2016) formally registered. Clinical trial registration could help to minimize bias in selective reporting and improve the validity and value of the scientific evidence (Angelis et al., 2006). Fourthly, intention to treat (ITT) analysis is a strategy to gather data as completely as possible on all randomized patients in line with their scheduled assessments (Lewis and Machin, 1993). Four RCTs ( -Geug et al., 2013;Kim et al., 2016;Lee et al., 2016) described whether they analyzed the data based on the ITT principle. Trials with incorrect or no ITT analysis may overestimate the results (Bondemark and Abdulraheem, 2017). Finally, various scales of fatigue were used as outcome measure in different trials. The evaluation of clinical efficacy rate is based on inconsistent scales, which restrict the validity and reliability. The quality of animal studies was considered to be moderate, suggesting that the results should be carefully interpreted. Fatigue can be divided into CFS and post-exercise fatigue in clinic. In the present analysis, all fatigue models are exhaustion, which may lead to ignore chronic process. In addition, we cannot neglect the contribution of Korea and other countries worldwide in the study of ginseng in treatment of fatigue. Due to the limitation of language, the present study was not included studies which language was not English or Chinese. We can increase international cooperation to overcome the linguistic limitation.

Implications
The evidence available from present study supported the routine use of PGC for fatigue, whereas the safety still needs more data because only five of eight studies reported. Given the gap between limitations of the primary RCTs and the quality of RCTs, we recommend that further design of the RCTs should consult the CONSORT statement (Moher et al., 2005), which offer a standard way for authors to design, conduct, analyze, and interpret, and to assess the validity of results.
Currently, there is no gold standard for measuring fatigue. Various scales were used to measure fatigue in different studies. Some scales have been calibrated, and some are homemade. Measurement of fatigue is challenging. Due to the wide range of conceptualizations of the problem and the concurrent development of questionnaires for many specific diseases, many questionnaires are used to measure fatigue. A comprehensive fatigue measurement, such as the Fatigue Severity Scale (Krupp et al., 1989), Piper Fatigue Scale (Piper et al., 1998), or FACIT-F (David et al., 2010), assesses the impact of fatigue on daily activities and its severity. In addition, short fatigue measurements such as the POMS-B fatigue subscale (Mcnairpm and Dropplemannl, 1992) and the 7-item Patient Reported Outcome Measurement Information System Cancer Fatigue Short Form (Cessna et al., 2016;Garcia et al., 2016) mainly assess severity of fatigue. The fatigue measurement precision with a comprehensive measure was greater than that with short fatigue measurement in evaluating moderate to severe fatigue, whereas the short fatigue measurement performed better in evaluating mild fatigue (Voshaar et al., 2015). Therefore, we should select the suitable fatigue measurements based on the research requirements.
Degree and scope of debilitating fatigue is a core component of health care where chronic diseases are receiving increasing attention. Current acute disease research models are not enough to solve chronic disorders affecting multiple regulatory systems and present complex constellations of symptoms. The identification of objective markers consistently associated with CFS is an important goal in relation to diagnosis and treatment, as the current case definitions are based entirely on physical signs and symptoms. Since the human body is an autonomous, fully integrated, and self-regulating system, it is not surprising that even localized muscle fatigue can present systemic biomarkers. There is a growing study devoted to understanding the biology of fatigue. Recognition of CFS biomarkers is an important part of this work. A complex construct of symptoms emerges from alterations and/ or dysfunctions in the nervous, endocrine, and immune systems. Biomarkers, distributing across these systems, constituted complex biological networks. The acquisition of biomarkers required a comprehensive biological network-based analysis of fatigue biology (Klimas et al., 2012). In addition, molecular aberrations observed in many CFS blood cell studies provided an opportunity to develop diagnostic analysis of blood samples. With the development of micro/nanofabrication, direct electrical detection of cellular and molecular properties, microfluidics, and artificial intelligence techniques, a nano-electronics blood-based assay have been developed, which can potentially establish diagnostic biomarkers and drug screening platform for CFS (Esfandyarpour et al., 2019).
Preclinical research is the key to convert preclinical data into clinical data (Ramirez et al., 2017). However, there is growing concern that poor experimental design and transparent reporting lead to frequent failure of translating preclinical discoveries into new therapies for human diseases (Hackam and Redelmeier, 2006). In present study, the quality of including animal studies was moderate. We recommend that further design of the studies should consult the ARRIVE guidelines (Kilkenny et al., 2012) and use appropriate animals, random allocation, model blinded induction, and outcomes blinded assessment to improve the accuracy of the results.
PGC acted through complex, multicompound, multitarget, and multipathway mechanisms in fatigue and might prove to be of great value in further clinical trials. The possible mechanisms of PGC for fatigue are summarized as follows: (1) Antioxidant stress: PGC passed through the injured membrane, improved the activity of GSH-Px, CAT, and SOD, scavenged hydroxyl radical, and reduced the excessive ROS, and thus preventing lipid oxidation and protecting the corpuscular membrane to reduce the release of LDH, MDA, and CK into the serum Zheng et al., 2017). Another mechanism might involve the nitric oxide pathway. NOS, a pro-oxidative enzyme, increased the production of toxic oxidant peroxynitrite. PGC decreased NOS and prevented peroxynitrite-induced mitochondrial dysfunction and lipid peroxidation (Ki Sung et al., 2006); (2) Regulation of carbohydrate metabolism: PGC increased the proportion of energy supplied by fat and promoted gluconeogenesis to improve hepatic glycogen storage. PGC enhanced the delivery of glucose by increasing capillary perfusion and plasma glucose concentration and increased the permeability of the muscle membrane of glucose to increase the muscle glucose uptake during exercise Ma et al., 2017). A reduced rate of hepatic and muscle glycogen break-down and a greater potential for fatty acid metabolism could maintain blood glucose level, and thus enhancing exercise capacity (Favier and Koubi, 1988); (3) Delaying the accumulation of metabolites: With the accumulation of BUN and BLA, the pH in muscle tissue and blood reduced, which could obstruct the transmission of excitation at neuromuscular junctions, reduce the maximum tension and sustainability of muscle tissue, and hinder the process of sugar supply. PGC increased the LDH activity and the glycogen levels and retarded the accumulation of BUN and BLA Delgado et al., 2019); (4) Promotion of mitochondrial function: SDH is a key enzyme associated with the regulation of the tricarboxylic acid cycle, catalyzing the synthesis of ATP. In addition, Na + -K + -ATPase and Ca 2+ -Mg 2+ -ATPase are crucial enzymes to degrade ATP. NRF-1 and TFAM are positive regulators of transcription. PGC improved SDH, Na + -K + -ATPase, and Ca 2+ Mg 2+ -ATPase activities. PGC enhanced the mRNA expression of NRF-1 and TFAM and increased the mtDNA content, and thereby enhancing mitochondrial function and producing more ATP for energy supplementation ; (5) Neuroprotection and antiapoptosis: PGC had a protective effect against MPTP-induced apoptosis in the mouse substantial nigra. This anti-apoptotic effect of PGC may be attributed to enhanced expression of Bcl-2 and Bcl-xl, reduced expression of bax and nitric oxide synthase, and inhibited activation of caspase-3 (Radad et al., 2006); (6) Regulation of neurotransmitter disorder: PGC decreased GABA and 5-HT levels, thereby increasing central nervous system excitability. PGC decreased the activity of acetylcholinesterase, maintained normal Ach and norepinephrine levels in cholinergic neurons, and enhanced the level of DA in the hippocampus .

CONCLUSIONS
The findings of present study demonstrated that PGC exerted antifatigue function, mainly through antioxidant stress, regulation of carbohydrate metabolism, delaying the accumulation of metabolites, promotion of mitochondrial function, neuroprotection and antiapoptosis, and regulation of neurotransmitter disorder. And the findings supported, at least to an extent, the use of PGC for fatigue.

AUTHOR CONTRIBUTIONS
Study conception and design: YL/G-QZ. Acquisition, analysis and/or interpretation of data: T-YJ/P-QR/H-YL/P-PZ/G-QZ/YL Final approval and overall responsibility for this published work: YL/G-QZ.