This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (17). We conducted a systematic electronic literature search in February 2021 in PubMed, Embase, Web of Science, and Cochrane Library databases. The search was limited to English. We used medical subject heading (MeSH and Emtree) terms combined with Boolean logical operators (for searching strategies, see Table S1).

Studies were included if they met the following criteria: 1) only RCT studies; 2) idiopathic infertility was diagnosed; 3) the patients have sperm parameter abnormalities; and 4) the primary outcomes have to include sperm motility, sperm concentration, sperm morphology, pregnancy rate, alone or in combination. Studies were excluded if they met the following criteria: 1) case–control, cross-sectional, or retrospective studies; 2) men with varicocele or other fertility-related diseases; 3) men with drug interventions; and 4) combined use of different types of antioxidants.

Two independent reviewers extracted the following data. Data extracted for the individual study included 1) general information related to the manuscript: first author, year of publication, and country. 2) Characteristics of the population: sample size, groups, age, and diagnosis. 3) Type of treatment, duration, and dosage. 4) Primary outcomes: sperm motility, sperm concentration, sperm morphology, and pregnancy rate. In cases that have multiple visits between these intervals, the longest one would be selected. We tried to make the included articles have similar characteristics to ensure the optimal transitivity of the NMA (18). Any dispute was resolved by consensus or consultation with a third reviewer.

The quality of RCTs was evaluated according to the Cochrane Collaboration’s tool (19), including random sequence generation, allocation concealment, blinding of participants, personnel and outcome assessment, incomplete outcome data, selective reporting, and other sources of bias.

All the analyses were performed using Stata statistical package version 14.1 software; all statistical analyses were conducted to calculate the weighted mean difference (WMD) for continuous data and odds ratio (OR) for dichotomous variables, together with the corresponding 95% CIs. For the traditional pairwise meta-analysis, the Mantel–Haenszel random-effects model was used (20). The heterogeneity in each pairwise comparison used I² statistics and was judged as low (<25%), moderate (25%–75%), or high (>75%). p-Values of <0.05 were regarded as statistically significant (21). The random-effects NMA was conducted as a frequentist approach (22). The model inconsistency must be evaluated because the presence of inconsistency is often regarded as a statistical manifestation of intransitivity (18). Both consistent and inconsistent models were analyzed and compared. Furthermore, the design-by-treatment interaction model and loop consistency were used to assess the inconsistency in the NMA. The loop-specific approach estimated inconsistency factors and their 95% CIs within every closed triangular or quadratic loop (23–25). The relative ranking of treatments was done using the surface under the cumulative ranking (SUCRA). The ranks for each outcome were also presented (26, 27). Additionally, publication bias of NMA and potential small-study effects were tested for using a comparison-adjusted funnel plot (28). Multiple sensitivity analyses were performed to address the robustness of our findings, including 1) exclusion of studies with a significant risk of bias, 2) exclusion of studies published before 2002, and 3) exclusion of sample size of less than 30.

According to the literature search and the inclusion criteria, we included 1,917 patients in 23 studies for meta-analysis (29–51) (Figure 1). The details are in Table 1. Those studies were published from 1998 to 2017 and evaluated 10 antioxidants including l-carnitine, l-carnitine+l-acetylcarnitine, coenzyme-Q10, ω-3 fatty acid, selenium, zinc, vitamin E+vitamin C, folic, and N-acetyl-cysteine. All the patients were diagnosed with idiopathic infertility with abnormal sperm parameters. Most studies included a two-arm study design, but 4 studies included a three-arm design. Of 23 studies, 2 were rated as low risk and 4 studies as high risk (Figure 2).

The results of the pairwise meta-analysis are shown in Table 2. We conducted the results of a pairwise meta-analysis that included two or more RCTs. In terms of sperm motility, LC (WMD 8.69% [95% CI: 1.54% to 15.84%]), LC+LAC (WMD 3.77% [95% CI: 1.02% to 6.52%]), Q10 (WMD 4.64% [95% CI: 4.00% to 5.28%]), ω-3 (WMD 4.33% [95% CI: 3.74% to 4.93%]), and Se (WMD 3.32% [95% CI: 2.65% to 3.99%]) had significantly increased sperm motility as compared with placebo. In terms of sperm concentration, the result showed that four antioxidants could improve sperm concentration compared with placebo—LC (WMD 3.65 × 10⁶/ml [95% CI: 2.42 to 4.89 × 10⁶/ml]), Q10 (WMD 5.74 × 10⁶/ml [95% CI: 4.52 to 6.95 × 10⁶/ml]), ω-3 (WMD 10.09 × 10⁶/ml [95% CI: 9.87 to 11.93 × 10⁶/ml]), and Se (WMD 3.91 × 10⁶/ml [95% CI: 2.85 to 5.44 × 10⁶/ml]). In terms of sperm morphology, LC (WMD 3.05% [95% CI: 2.60% to 3.50%]), CoQ10 (WMD 1.74% [95% CI: 1.06% to 2.42%]), and ω-3 (WMD 0.46% [95% CI: 0.22% to 0.71%]) were shown to have treatment effects with statistical significance. In terms of pregnancy rate, 12 studies reported the pregnancy rate, but no significant differences in the pregnancy rate were observed between antioxidants and placebo.

There is no evidence of publication bias in comparison-adjusted funnel plots. The funnel plots are shown in Figure 5. In terms of sensitivity, excluding published before 2002 and smaller studies also showed no obvious change in the primary outcomes. So it showed the stability of the results. The results are shown in Table S3.

LC, LC+LCA, CoQ10, and ω-3 showed beneficial effects on sperm motility as compared with placebo. Previous meta-analyses conducted by Buhling (52) and Zhang (53) also demonstrated that carnitine had a favorable effect on sperm motility as compared with placebo. Unlike previous studies’ previous pairwise meta-analyses, our NMA demonstrated that LC was the most effective treatment to increase sperm motility, followed by CoQ10. On the other hand, basic research has been reported that LC and LAC participate in the transportation of long-chain fatty acids to the mitochondrial matrix for β-oxidation, providing energy for β-oxidation of spermatozoa in the epididymis (54). So LC and LAC play a vital role in sperm motility, spermatogenic process, and maturation by enhancing energy metabolism (55). It seems to explain why LC has a better effect on sperm motility. However, some studies had an unclear randomization method and a low number of patients. So the results should be interpreted with caution.

Compared with placebo, ω-3, CoQ10, Se, and LC had significantly increased sperm concentration. Our NAM demonstrated that ω-3 had the highest probability of being the most effective treatment to increase sperm concentration, followed by CoQ10. Recently, basic research demonstrated that ω-3 fatty acids have anti-inflammatory and antioxidant properties, potentially protecting the composition and function of cell membranes. Furthermore, the successful fertilization of spermatozoa is related to the lipid composition of the spermatozoa membrane (56, 57). Consistent with this finding, the previous meta-analysis also showed that ω-3 fatty acids had a positive effect on sperm concentration compared with placebo (58). However, there are few reports on the influence of ω-3 on pregnancy rate, and no major side effects were reported from the included studies. So more RCTs conducted in large samples of participants are needed to verify the therapeutic effects of ω-3 on pregnancy rate and side effects.

The NMA demonstrated that CoQ10 and ω-3 were near the critical statistical significance as compared with placebo. But only LC was shown to have treatment effects with statistical significance. Contrastingly, our pairwise meta-analysis demonstrated that ω-3 and CoQ10 were shown to have treatment effects with statistical significance. The previous meta-analysis conducted by Salas (59) also demonstrated that ω-3 and CoQ10 had a favorable effect on sperm morphology as compared with placebo. The controversy of this result might be related to the lack of study and the differences in data after mixed comparison in the NMA. So we need more RCTs to verify the therapeutic effects on the morphology of ω-3 and CoQ10. However, there is no doubt that LC has the best therapeutic effect on sperm morphology.

The most objective outcome to show the effect of treatment on male fertility must be the pregnancy rate. However, only 12 RCTs reported the pregnancy rate. There is no statistical significance between treatment and placebo. Most studies have reported effects on sperm parameters but have not described the pregnancy rate in detail. On the other hand, the shorter follow-up time would result in lower positive rates. We must point out that “fertility” also depends on the fertility status of the female partner. It would also clearly influence the outcome of medical intervention in the male partner, although the previous meta-analysis showed that supplementation with antioxidants has a positive effect on pregnancy rate as compared with placebo (60). However, due to the lack of data in each study, the results are not applicable to idiopathic male infertility.

Although CoQ10 was not ranked first in four aspects of sperm parameters in this study, CoQ10 has an obvious therapeutic effect on sperm motility and sperm concentration. A previous meta-analysis study also concludes that CoQ10 has a profound effect on sperm motility and concentration (61). In the electron-transport system, CoQ10 is an antioxidant molecule that plays a vital role, and CoQ10 could inhibit the formation of organic peroxides in sperm to reduce sperm cell OS (62, 63). Based on our result, CoQ10 could be a good choice for idiopathic male infertility.

It is worth emphasizing that previous RCTs and meta-analyses did not analyze the optimal dosage and medication time. Until now, the exact dosages and regimens have not been clearly defined (64). Recently, a study has shown that excessive intake of antioxidants can change the oxidation–reduction equilibrium into reductive stress, and it has the same adverse effect on sperm quality as OS (65). In our study, the follow-up period was from 3 to 8 months, and the dosage of the same type of antioxidants was roughly the same. Compared with placebo, some antioxidants such as zinc, vitamin E+vitamin C, and folic were of no statistical significance. But we do not know whether the reductive stress was caused by a high amount/dose of antioxidants or lack of follow-up time. On the other hand, some studies have reported that hormones and SERMs increased sperm quality (66). We need more experimental studies on these interventions with significant effect sizes so that a better treatment could be planned to improve the outcome.

Several limitations of our NMA should be stated. The internal effectiveness of NMA mainly relies on the realization of the required assumptions, like heterogeneity, inconsistency, and transitivity (67). First, although the results of inconsistency factors were not found in the loops of treatment efficacy, the global inconsistency was existing in the outcome of sperm motility. Second, we found that some treatments had high statistical heterogeneity based on I². It could be caused by methodological heterogeneity (clinical measurement tool and point of evaluation) and clinical heterogeneity (age, duration of infertility, or dosage). So considering the heterogeneity between studies, we used the random-effects model to pool the pairwise estimates. Third, many treatment pairs include two studies or at the most a maximum of three studies, which were not sufficient to correctly assess the transitivity assumption. So we are looking forward to getting more clinical data in the future. Furthermore, the applicability of our findings might be limited, because the NMA was based on indirect comparisons but not direct comparisons. At last, we have included the main outcomes of sperm motility, sperm morphology, sperm concentration, and pregnancy rate. Because of the lack of data in each study, we could not include some outcomes like forwarding sperm motility and sperm volume. Therefore, we have to be cautious when evaluating the outcomes of different treatments.

We also hope that this article will give some enlightenment to the focus in our future research. First of all, some RCTs have gradually begun to use combinations of different types of antioxidants, for example, using the mixture of selenium+zinc+vitamin E or LC+CoQ10+vitamin C+zinc (68, 69). Those combinations were also more efficacious than placebo in sperm quality parameters. However, due to the lack of research literature and difficulty in distinguishing the effects of combinations of various antioxidants, those studies were not included. Therefore, in the future, we need more RCTs to verify the therapeutic effects of combining different types of antioxidants.