We followed the guidelines delineated in the latest version of the PRISMA 2020 guidelines (Supplementary Table S1) to conduct this meta-analysis (Page et al., 2021a). This study, which was registered in INPLASY with the registration number INPLASY202210113 (Tsai and Chang, 2022), did not require ethics review board approval or participant informed consent.

Two authors (I-CT and K-VC) conducted independent electronic searches in the PubMed, Embase, Cochrane CENTRAL, Web of Science, and ClinicalTrials.gov databases using the following keywords (“Q10” OR “Q 10” OR “CoQ10” OR “Coenzyme Q10” OR “ubiquinol-10” OR “ubiquinol” OR “ubiquinone”) AND (“fatigue” OR “chronic fatigue syndrome” OR “tiredness”). The search was conducted from the inception of each database (i.e., the earliest record) to the date of the database search (16 January 2022). We note that ubiquinone is the fully oxidized form of CoQ10, and ubiquinol is the reduced form. These two forms are continually interconverted in the body and have similar bioactivities (Mantle and Dybring, 2020). The detailed search strategy for this systematic review and meta-analysis is provided in the Supplementary Material (Supplementary Table S2).

Initially, the two authors responsible for conducting this search screened the identified titles and abstracts for eligibility through a consensus process. The PubMed and EMBASE databases were thoroughly scrutinized for any potentially eligible trials. We also checked the reference lists of an identified review article (Mehrabani et al., 2019) and performed additional manual searches. A third reviewer and study author (P-TT) was consulted for situations in which the two aforementioned authors could not reach a consensus. No language restrictions were applied to this search.

The PICO (population, intervention, comparison, outcome) setting of the current meta-analysis was as follows: P: human participants; I: CoQ10 supplementation; C: placebo; and O: changes in fatigue symptom scores.

The following inclusion criteria were used: 1) randomized controlled trials (RCTs) enrolling human participants, 2) RCTs investigating the quantitative evaluation of fatigue symptoms before and after CoQ10 supplementation, 3) placebo-controlled trials (without age or treatment duration limitations), and 4) trials with available data for pre- and post-intervention fatigue assessments or evaluations of changes in fatigue scores. Open-label studies were also included in this meta-analysis as recent studies have found that open-label placebos had similar efficacy to double-blind placebos (Lembo et al., 2021; von Wernsdorff et al., 2021).

The exclusion criteria for this review and meta-analysis were as follows: 1) non-RCTs, 2) studies focusing on athletic muscle exhaustion rather than generalized fatigue, 3) studies lacking a placebo-controlled group, 4) studies lacking quantitative assessments of fatigue, and 5) studies enrolling participants that overlapped with a previously published trial.

To investigate the methodological quality of the evaluated studies, we used the Cochrane risk of bias tool for randomized trials (version 2, RoB 2, London, United Kingdom), which consists of six main items for evaluating study quality: the randomization process, intervention adherence, missing outcome data, outcome measurement, selective reporting, and the overall risk of bias (Sterne et al., 2019).

In the intervention adherence section of the RoB 2, there are two options presented for literature assessment: intention-to-treat (intervention assignment) and per-protocol (intervention adherence). In this meta-analysis, we chose the per-protocol evaluation (Sterne et al., 2019) since it fits best with the design of our included studies.

The primary outcomes evaluated in this investigation were changes in fatigue scores following CoQ10 supplementation or placebo regimens. We also examined the validity and appropriateness of the fatigue scale used in each trial (Fisk et al., 1994; Chandran et al., 2007; Fukuda et al., 2008). If there was more than one scoring system for fatigue evaluation implemented in a single trial, the index test included in the meta-analysis was decided by author consensus (I-CT and K-VC).

The secondary outcome evaluated in this investigation was the treatment-associated adverse event rate. For cells with zero events, zero was replaced by 0.5 to enable calculation (Deeks et al., 2021). The aforementioned outcomes were quantified using odds ratios.

Two independent authors (I-CT and K-VC) extracted data from the evaluated studies, including demographic data, study design parameters, details of the administered CoQ10 and placebo regimens, and the primary and secondary outcome values. To avoid misinterpretation, the evaluators paid special attention to the effect direction of the scale used in each trial. In situations where data were unavailable within the published articles, we contacted the corresponding authors to obtain the original data.

Data extraction and conversion as well as merging the results from various study arms using different CoQ10 dosages were processed according to the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions and the associated medical literature (Hozo et al., 2005; Higgins et al., 2021a; Higgins et al., 2021b). If post-treatment data were available at multiple time points, we extracted the outcome reported at the end of the intervention for statistical analysis. For crossover studies, we included only the first study interval to avoid carry-over effects (Higgins et al., 2021a).

Because of the heterogeneity of the target populations within the enrolled studies, the current meta-analysis was conducted with a random-effects model (Borenstein et al., 2009) implemented using Comprehensive Meta-Analysis software (version 3, Biostat, Englewood, NJ, United States). A two-tailed p value of less than 0.05 was considered statistically significant.

We used Hedges’ g and 95% confidence intervals (CIs) to quantify the primary study outcomes (i.e., changes in fatigue scores). Hedges’ g values of 0.2, 0.5, and 0.8 were considered small, moderate, and large effect sizes, respectively (Hedges, 1981). Odds ratios (ORs) and their associated 95% CIs were evaluated to investigate secondary outcomes (i.e., treatment-associated adverse event rates).

I ² and Cochran’s Q statistics were also examined to evaluate the degree of heterogeneity across studies. I ² values of 25, 50, and 75% were considered low, moderate, and high heterogeneity, respectively (Higgins et al., 2003). We likewise performed subgroup analyses based on disease and CoQ10 formulations. Meta-regression analyses regarding the treatment effects of daily CoQ10 doses as well as evaluations of various treatment durations were conducted to determine whether the fatigue-alleviating effects of CoQ10 correlated with the aforementioned parameters.

To confirm the robustness of this meta-analysis, sensitivity analyses were performed using the one-study removal method to determine whether there was a statistically significant change in the summary effect size after removing a particular trial from the analysis (Deeks et al., 2021).

Potential publication bias was evaluated using guidelines set forth by the Cochrane Handbook for Systematic Reviews of Interventions (Page et al., 2021b). Funnel plots were generated and visually inspected. Egger’s regression tests were implemented when 10 or more datasets were available.

The PRISMA flowchart for the literature search is shown in Figure 1. After removing duplicate articles and excluding non-relevant articles by an examination of titles and abstracts, we ultimately included 13 RCTs in the final analysis (Berman et al., 2004; Lee et al., 2011; Cordero et al., 2013; Lesser et al., 2013; Castro-Marrero et al., 2015; Fukuda et al., 2015; Peel et al., 2015; Sanoobar et al., 2016; Di Pierro et al., 2017; Morikawa et al., 2019; Mizuno et al., 2020; Mousavi et al., 2020; Castro-Marrero et al., 2021). The articles excluded in the final stage and the reasons for exclusion are listed in Supplementary Table S3 (Singh et al., 2003; Langsjoen et al., 2005; Kumar et al., 2007; Mizuno et al., 2008; Gökbel et al., 2010; Cordero et al., 2012a; Cordero et al., 2012b; Fedacko et al., 2013; Gharahdaghi et al., 2013; Miyamae et al., 2013; Castro-Marrero et al., 2016; Fukuda et al., 2016; Iwase et al., 2016; Menon et al., 2017; Langsjoen et al., 2019; Moccia et al., 2019; Negro et al., 2019; Gomez-Centeno et al., 2020; Schweiger et al., 2020; Suzuki et al., 2021). Details of data extraction from included randomized controlled trials are summarized in Supplementary Table S4.

The 13 eligible RCTs encompassed a total of 1,126 participants with a mean age of 49.3 ± 12.6 (standard deviation) years, of whom 25.6% (n = 288) were male. The study duration ranged from four (Mousavi et al., 2020) to 24 weeks (Lesser et al., 2013). Subject diagnoses included chronic fatigue syndrome (Castro-Marrero et al., 2015; Castro-Marrero et al., 2021), fibromyalgia (Cordero et al., 2013; Di Pierro et al., 2017), end-stage heart failure (Berman et al., 2004), obesity (Lee et al., 2011), breast cancer (with patients undergoing chemotherapy) (Lesser et al., 2013), end-stage renal disease (Fukuda et al., 2015), poliomyelitis (Peel et al., 2015), and multiple sclerosis (Sanoobar et al., 2016), and we also evaluated healthy subjects (Morikawa et al., 2019) and healthy individuals with fatigue (Mizuno et al., 2020; Mousavi et al., 2020). The details of the retrieved trials are summarized in Table 1.

Three studies evaluated compounds mixed with CoQ10. More specifically, CoQ10 with a reduced form of nicotinamide adenine dinucleotide (NADH) was evaluated in two of these trials (Castro-Marrero et al., 2015; Castro-Marrero et al., 2021) and CoQ10 in a multi-vitamin nutritional drink was evaluated in the other trial (Fukuda et al., 2015). Ten studies evaluated CoQ10 only (Berman et al., 2004; Lee et al., 2011; Cordero et al., 2013; Lesser et al., 2013; Peel et al., 2015; Sanoobar et al., 2016; Di Pierro et al., 2017; Morikawa et al., 2019; Mizuno et al., 2020; Mousavi et al., 2020). The intervention details, tools for fatigue assessment, adverse events, and study withdrawals are summarized in Table 2.

With respect to the overall methodological quality of the included studies, we found that 46.2% of the evaluated studies had a low risk of bias, 53.8% had some risk of bias, and 0% had a high risk of bias (Figure 2). In a detailed assessment, seven studies (Lee et al., 2011; Cordero et al., 2013; Lesser et al., 2013; Castro-Marrero et al., 2015; Sanoobar et al., 2016; Morikawa et al., 2019; Mizuno et al., 2020) were rated as having “some” risk of bias in the randomization process because they did not reveal details of the allocation concealment. One study (Lee et al., 2011) was rated as having “some” risk of bias with regard to missing outcome data, since the study excluded five subjects from the CoQ10 group due to an unchanged level of serum CoQ10 after supplementation. One study (Di Pierro et al., 2017) was rated as having “some” risk of bias in the outcome measurement because it was an open-label study and the participants were aware of the interventions they received. The details of the risk of bias assessment are summarized in Table 3.

In the combined 13 trials (Figure 3), CoQ10 demonstrated a statistically significant reduction with regard to fatigue symptomology (Hedges’ g = −0.398, 95% CI = −0.641 to −0.155, p = 0.001, I ² = 69.5%). However, moderate-to-high heterogeneity was observed. A sensitivity analysis was performed using the one-study removal method. The results showed a consistently statistically significant effect of CoQ10 on fatigue reduction. The summary effect sizes did not change the statistical significance of these findings when any of the included studies were removed (Figure 4).

We then divided the included trials into two subgroups: healthy participants (Morikawa et al., 2019; Mizuno et al., 2020; Mousavi et al., 2020) and patients with a fatigue-associated disease (fibromyalgia (Cordero et al., 2013; Di Pierro et al., 2017), chronic fatigue syndrome (Castro-Marrero et al., 2015; Castro-Marrero et al., 2021), heart failure (Berman et al., 2004), obesity (Lee et al., 2011), breast cancer (Lesser et al., 2013), end-stage renal disease (Fukuda et al., 2015), poliomyelitis (Peel et al., 2015), multiple sclerosis (Sanoobar et al., 2016)). The direction of association between the use of CoQ10 and fatigue assessment was consistent in the subgroups of healthy participants (Hedges’ g = −0.351, 95% CI = −0.756 to 0.053, p = 0.089) and in patients with disease (Hedges’ g = −0.433, 95% CI = −0.732 to −0.133, p = 0.005), with overlapping 95% CIs (Figure 5).

Another subgroup analysis was performed according to differences in the evaluated CoQ10 regimens. The group that used only CoQ10 showed a statistically significant fatigue reduction after supplementation (Hedges’ g = −0.552, 95% CI = −0.863 to −0.241, and p = 0.001). However, the effect of fatigue reduction in the group using CoQ10 compounds (Figure 6) was trivial and statistically insignificant as compared with the placebo group (Hedges’ g = −0.028, 95% CI = −0.225 to 0.170, p = 0.781).

Meta-regression was performed to examine whether daily CoQ10 dose and treatment duration could modify effects on fatigue reduction. Both daily doses (coefficient = −0.0017 per mg, p < 0.001) and treatment duration (coefficient = −0.0042 per day, p = 0.007) correlated with increased fatigue reduction (Figures 7, 8). The funnel plot of the 13 included trials showed some asymmetry in effect size distributions (Supplementary Figure S1). Egger’s regression test showed a p-value of 0.008, indicating potential publication bias.

Among the 602 participants treated with CoQ10, only one subject presented with an adverse event (i.e., gastrointestinal upset) and was thus withdrawn from the study conducted by Berman et al. (Berman et al., 2004). The meta-analysis of treatment-associated adverse event rates (Supplementary Figure S2) showed no between-group differences (odds ratio [OR] = 1.05, 95% CI = 0.36 to 3.08, p = 0.933, I ² < 0.01%).