Safety Evaluation of α-Lipoic Acid Supplementation: A Systematic Review and Meta-Analysis of Randomized Placebo-Controlled Clinical Studies

Alpha-lipoic acid (ALA) is a natural short-chain fatty acid that has attracted great attention in recent years as an antioxidant molecule. However, some concerns have been recently raised regarding its safety profile. To address the issue, we aimed to assess ALA safety profile through a systematic review of the literature and a meta-analysis of the available randomized placebo-controlled clinical studies. The literature search included EMBASE, PubMed Medline, SCOPUS, Google Scholar, and ISI Web of Science by Clarivate databases up to 15th August 2020. Data were pooled from 71 clinical studies, comprising 155 treatment arms, which included 4749 subjects with 2558 subjects treated with ALA and 2294 assigned to placebo. A meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of any treatment-emergent adverse event (all p > 0.05). ALA supplementation was safe, even in subsets of studies categorized according to smoking habit, cardiovascular disease, presence of diabetes, pregnancy status, neurological disorders, rheumatic affections, severe renal impairment, and status of children/adolescents at baseline.


Introduction
Alpha-lipoic acid (1, 2-dithiolane-3-pentanoic acid; ALA) or thioctic acid is a natural short-chain fatty acid that has attracted great attention in recent years as an antioxidant molecule, being largely used worldwide as a dietary supplement [1].
Previous investigations revealed that ALA can affect central and peripheral modulation of 5 -adenosine-monophosphate-activated protein kinase. Furthermore, it activates peroxisome proliferator-activated receptor (PPAR) alpha and gamma (PPAR-γ), modulates PPAR-regulated genes and upregulates the expression of PPAR-γ messenger ribonucleic acid (mRNA) and other proteins in the cardiac tissue and aorta smooth muscle [2,3]. Hence, ALA antioxidant activity is potentially able to promote weight loss and blood pressure control and ameliorate atherogenic dyslipidemia and insulin resistance [3]. For example, in obese patients with non-alcoholic fatty liver disease (NAFLD), ALA supplementation was shown to reduce adipokine concentrations and improve liver steatosis grade [4,5]. However, some concerns have been recently raised regarding ALA safety profile, after some reports suggesting a direct causal link between its use and insulin autoimmune syndrome (IAS, also known as Hirata's disease) due to its sulfhydryl group [6]. Indeed, in about 50% of cases, IAS development is associated with drugs or dietary supplement containing a sulphur or sulfhydryl group. These cases are closely related to certain specific antigens of the major histocompatibility complex (MHC), which are more common in populations where IAS incidence is higher [7]. It is hypothesised that ALA might cause the development of antibodies to insulin and lead to a hypoglycaemic syndrome in predisposed subjects, even though evidence are inconclusive [8].
In a recent study that performed a preliminary analysis of spontaneous reports of suspected adverse reactions (ARs), ALA-containing natural products have also been associated with skin and gastrointestinal disorders, such as urticaria and abdominal pain [9].
To address safety issues related to ALA supplementation, we aimed to perform a systematic review of the literature and a meta-analysis of the available randomized placebo-controlled clinical trials.

Materials and Methods
The study was designed according to guidelines of the 2009 preferred reporting items for systematic reviews and meta-analysis (PRISMA) statement [10], and was registered in the PROSPERO database (Registration number CRD42020159028).
Due to the study design, neither Institutional Review Board (IRB) approval, nor patient informed consent were required. PRISMA Checklist was reported in supplementary file A. versus a single control group, the number of subjects in the control group was divided by the required comparisons [15].
To reduce the risk of bias due to effect dilution, the meta-analysis was performed considering per-protocol (PP) population.
Studies' findings were combined using a fixed-effect model since the low level of inter-study heterogeneity, which was quantitatively assessed using the Higgins index (I 2 ) [16]. Effect sizes were expressed as odds ratio (OR) and 95% confidence interval (95% CI) [17]. Finally, sensitivity analysis was conducted to account for the risk of bias. A leave-one-out method was used (i.e., one study was removed at a time and the analysis was repeated) [18].
Two-sided p-values < 0.05 were considered as statistically significant for all tests.

Additional Analysis
Subgroup analyses were carried out by presence of smoking habit, pregnancy, CV disease, diabetes, rheumatic disorders, neurological disorders, severe renal impairment, and status of children/adolescent at baseline.

Publication Biases
Potential publication biases were explored using visual inspection of Begg's funnel plot asymmetry, Begg's rank correlation test, and Egger's weighted regression test [19]. Two-sided p-values < 0.05 were considered statistically significant for the tests.

Flow and Characteristics of the Included Studies
After database searches performed strictly according to inclusion and exclusion criteria, 962 published articles were identified, and their abstracts reviewed. Of these, 359 did not report original data. Furthermore, 393 articles were excluded because they did not meet the inclusion criteria. Thus, 210 articles were carefully assessed and reviewed. Additional 139 papers were excluded due to being pre-print papers (n = 2), study protocols (n = 6), reporting data from studies lacking of an appropriate placebo-controlled design for the supplementation (n = 64), lacking of randomisation (n = 5), testing the acute effect of ALA supplementation (n = 7), testing ALA supplementation combined in nutraceutical compounds (n = 27), testing intravenous treatment with ALA (n = 11), testing topical treatment with ALA (n = 4), lacking sufficient information about the nature of the adverse events (n = 9), or reporting data overlapped with other publications (n = 4) (Supplementary file B). Finally, 71 studies were eligible and included in the systematic review . The study selection process is shown in Figure 1.
Data were pooled from 71 randomized placebo-controlled clinical studies, comprising 155 treatment arms (82 active arms and 73 control arms). The studies included 4749 subjects, with 2558 receiving treatment with ALA and 2294 subjects assigned to placebo. For reasons independent of the tested supplementation (i.e., withdrawal of informed consent and personal problems), 510 subjects prematurely terminated the trials in which they were enrolled. Then, the meta-analysis was performed considering the other subjects (i.e., PP population). Data were pooled from 71 randomized placebo-controlled clinical studies, comprising 155 treatment arms (82 active arms and 73 control arms). The studies included 4749 subjects, with 2558 receiving treatment with ALA and 2294 subjects assigned to placebo. For reasons independent of the tested supplementation (i.e., withdrawal of informed consent and personal problems), 510 subjects prematurely terminated the trials in which they were enrolled. Then, the meta-analysis was performed considering the other subjects (i.e., PP population).
Eligible studies were published between 1982 and 2020 and were conducted in different locations across all continents. Follow-up periods ranged between 8 days and 4 years and several ALA regimens were tested. Selected clinical trials were designed with cross-over or parallel-group and enrolled pregnant women with gestational diabetes, children and/or adolescent, overall healthy subjects or subjects with minor or major underlying diseases (e.g., diabetes, CVD, rheumatic affections, neurological disorders, severe renal impairment).
The main characteristics of the evaluated studies are summarized in Table 1. Eligible studies were published between 1982 and 2020 and were conducted in different locations across all continents. Follow-up periods ranged between 8 days and 4 years and several ALA regimens were tested. Selected clinical trials were designed with cross-over or parallel-group and enrolled pregnant women with gestational diabetes, children and/or adolescent, overall healthy subjects or subjects with minor or major underlying diseases (e.g., diabetes, CVD, rheumatic affections, neurological disorders, severe renal impairment).
The main characteristics of the evaluated studies are summarized in Table 1.

Risk of Bias Assessment
Almost all of the included studies were characterized by sufficient information regarding sequence generation, allocation concealment, personal and outcome assessments, incomplete outcome data, and selective outcome reporting. Details of the quality of bias assessment are reported in Table 2.    Jacob, 1999 [44] L   The quality of evidence for each outcome across all the studies was considered high in accordance with the GRADE approach.

Hypoglycaemic Episodes
Symptoms defined as 'similar to hypoglycaemic episodes' were reported only by Jacob et al. and were exclusively experienced by subjects randomized to placebo. Authors did not report if an attempt for treatment rechallenging was made during the trial [44].

Gastrointestinal AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of gastrointestinal AEs (OR = 1.32, 95% CI 0.97 to 1.78; p = 0.073; I 2 = 0%) ( Figure 2). The finding was robust in the leave-one-out sensitivity analysis ( Figure S1).  Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S2). This asymmetry was imputed to eight potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 1.12 (95% CI 0.84 to 1.49). Egger's linear regression and Begg's rank correlation confirmed the presence of publication bias for the analysis (p < 0.05).

Neurological AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of neurological AEs (OR = 1.53, 95% CI 0.88 to 2.63; p = 0.129; I 2 = 0%) ( Figure  3). The finding was robust in the leave-one-out sensitivity analysis ( Figure S3). Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S2). This asymmetry was imputed to eight potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 1.12 (95% CI 0.84 to 1.49). Egger's linear regression and Begg's rank correlation confirmed the presence of publication bias for the analysis (p < 0.05).

Neurological AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of neurological AEs (OR = 1.53, 95% CI 0.88 to 2.63; p = 0.129; I 2 = 0%) (Figure 3). The finding was robust in the leave-one-out sensitivity analysis ( Figure S3).
Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S4). This asymmetry was imputed to 4 potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 1.26 (95% CI 0.76 to 2.10). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).  Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S4). This asymmetry was imputed to 4 potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 1.26 (95% CI 0.76 to 2.10). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

Psychiatric disorders
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of psychiatric disorders (OR = 1.13, 95% CI 0.64 to 1.99; p = 0.668; I 2 = 0%) ( Figure 4). The finding was robust in the leave-one-out sensitivity analysis ( Figure S5).

Study name
Statistics for each study Odds ratio and 95% CI

Psychiatric Disorders
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of psychiatric disorders (OR = 1.13, 95% CI 0.64 to 1.99; p = 0.668; I 2 = 0%) ( Figure 4). The finding was robust in the leave-one-out sensitivity analysis ( Figure S5).

Study name
Statistics for each study Odds ratio and 95% CI  Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S4). This asymmetry was imputed to 4 potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 1.26 (95% CI 0.76 to 2.10). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

Psychiatric disorders
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of psychiatric disorders (OR = 1.13, 95% CI 0.64 to 1.99; p = 0.668; I 2 = 0%) ( Figure 4). The finding was robust in the leave-one-out sensitivity analysis ( Figure S5).

Study name
Statistics for each study Odds ratio and 95% CI   Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S6). This asymmetry was imputed to two potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 1.01 (95% CI 0.59 to 1.75). Egger's linear regression confirmed the presence of publication bias for the analysis (p < 0.01), though Begg's rank correlation did not.

Musculoskeletal AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of musculoskeletal AEs (OR = 0.76, 95% CI 0.22 to 2.64; p = 0.666; I 2 = 0%) ( Figure 5). The finding was robust in the leave-one-out sensitivity analysis ( Figure S7). presence of publication bias for the analysis (p < 0.01), though Begg's rank correlation did not.

Musculoskeletal AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of musculoskeletal AEs (OR = 0.76, 95% CI 0.22 to 2.64; p = 0.666; I 2 = 0%) ( Figure 5). The finding was robust in the leave-one-out sensitivity analysis ( Figure S7).  Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S8). This asymmetry was imputed to 2 potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 0.50 (95% CI 0.17 to 1.51). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

Skin AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of skin AEs (OR = 1.13, 95% CI 0.82 to 1.56; p = 0.469; I 2 = 33.6%) (Figure 6). The finding was robust in the leave-one-out sensitivity analysis ( Figure S9). Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S8). This asymmetry was imputed to 2 potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 0.50 (95% CI 0.17 to 1.51). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

Skin AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of skin AEs (OR = 1.13, 95% CI 0.82 to 1.56; p = 0.469; I 2 = 33.6%) (Figure 6). The finding was robust in the leave-one-out sensitivity analysis ( Figure S9).  Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S10). This asymmetry was imputed to four potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 0.92 (95% CI 0.68 to 1.24). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

Infections
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of infections (OR = 0.93, 95% CI 0.18 to 4.65; p = 0.925; I 2 = 0%) (Figure 7). The finding was robust in the leave-one-out sensitivity analysis ( Figure S11).

Study name
Statistics for each study Odds ratio and 95% CI Odds Lower Upper ratio limit limit Z-Value p-Value Infections Figure 6. Forest plot for the risk of skin AEs following ALA supplementation versus placebo.
Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S10). This asymmetry was imputed to four potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 0.92 (95% CI 0.68 to 1.24). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

Infections
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of infections (OR = 0.93, 95% CI 0.18 to 4.65; p = 0.925; I 2 = 0%) (Figure 7). The finding was robust in the leave-one-out sensitivity analysis ( Figure S11). > 0.05 for both tests).

Infections
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of infections (OR = 0.93, 95% CI 0.18 to 4.65; p = 0.925; I 2 = 0%) (Figure 7). The finding was robust in the leave-one-out sensitivity analysis ( Figure S11).

Study name
Statistics for each study Odds ratio and 95% CI Odds Lower Upper ratio limit limit Z-Value p-Value Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S12). This asymmetry was imputed to two potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 0.31 (95% CI 0.08 to 1.13). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests). Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S12). This asymmetry was imputed to two potentially missing studies on the left-side of the plot, which reduced the estimated effect size to 0.31 (95% CI 0.08 to 1.13). However, neither Egger's linear regression nor Begg's rank correlation confirmed the presence of publication bias for the analysis (p > 0.05 for both tests).

CV System AEs
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of CV system AEs (OR = 1.25, 95% CI 0.84 to 1.85; p = 0.276; I 2 = 15.8%) (Figure 8). The finding was robust in the leave-one-out sensitivity analysis ( Figure S13). Antioxidants 2020, 9, x FOR PEER REVIEW 22 Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of CV system AEs (OR = 1.25, 95% CI 0.84 to 1.85; p = 0.276; I 2 = 15.8%) ( Figure  8). The finding was robust in the leave-one-out sensitivity analysis ( Figure S13).

Study name
Statistics for each study Odds ratio and 95% CI  Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S14). This asymmetry was imputed to three potentially missing studies on the right-side of the plot, which increased the estimated effect size to 1.40 (95% CI 0.95 to 2.05). Egger's linear regression confirmed the presence of publication bias for the analysis (p < 0.01), though Begg's rank correlation did not.

Hospitalisation
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of hospitalisation (OR = 5.66, 95% CI 0.64 to 49.85; p = 0.119; I 2 = 0%) (Figure 9). The finding was robust in the leave-one-out sensitivity analysis ( Figure S15).

Study name
Statistics for each study Odds ratio and 95% CI Hospitalization Figure 8. Forest plot for the risk of CV system AEs following ALA supplementation versus placebo.
Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S14). This asymmetry was imputed to three potentially missing studies on the right-side of the plot, which increased the estimated effect size to 1.40 (95% CI 0.95 to 2.05). Egger's linear regression confirmed the presence of publication bias for the analysis (p < 0.01), though Begg's rank correlation did not.

Hospitalisation
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of hospitalisation (OR = 5.66, 95% CI 0.64 to 49.85; p = 0.119; I 2 = 0%) (Figure 9). The finding was robust in the leave-one-out sensitivity analysis ( Figure S15). the presence of publication bias for the analysis (p < 0.01), though Begg's rank correlation did not.

Hospitalisation
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of hospitalisation (OR = 5.66, 95% CI 0.64 to 49.85; p = 0.119; I 2 = 0%) (Figure 9). The finding was robust in the leave-one-out sensitivity analysis ( Figure S15).

Study name
Statistics for each study Odds ratio and 95% CI  Figure 9. Forest plot for the risk of hospitalisation following ALA supplementation versus placebo.

Death
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of death (OR = 0.56, 95% CI 0.21 to 1.48; p = 0.242; I 2 = 0%) ( Figure 10). The finding was robust in the leave-one-out sensitivity analysis ( Figure S16).

Death
Meta-analysis of extracted data suggested that supplementation with ALA was not associated with an increased risk of death (OR = 0.56, 95% CI 0.21 to 1.48; p = 0.242; I 2 = 0%) ( Figure 10). The finding was robust in the leave-one-out sensitivity analysis ( Figure S16). Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S17). This asymmetry was imputed to three potentially missing studies on the right-side of the plot, which increased the estimated effect size to 0.71 (95% CI 0.31 to 1.64). Egger's linear regression correlation confirmed the presence of publication bias for the analysis (p = 0.03), though Begg's rank correlation did not.

Additional Analyses
Supplementation with ALA was not associated with a significant increased risk of any AE in subsets of studies classified by smoking habit, CV disease, diabetes, pregnancy, neurological disorders, rheumatic affections, and severe renal impairment at baseline (Table 3). Furthermore, ALA supplementation was safe in children ( Table 3). The findings were robust in the leave-one-out sensitivity analysis. Visually, the funnel plot of standard error by log OR was slightly asymmetric ( Figure S17). This asymmetry was imputed to three potentially missing studies on the right-side of the plot, which increased the estimated effect size to 0.71 (95% CI 0.31 to 1.64). Egger's linear regression correlation confirmed the presence of publication bias for the analysis (p = 0.03), though Begg's rank correlation did not.

Additional Analyses
Supplementation with ALA was not associated with a significant increased risk of any AE in subsets of studies classified by smoking habit, CV disease, diabetes, pregnancy, neurological disorders, rheumatic affections, and severe renal impairment at baseline (Table 3). Furthermore, ALA supplementation was safe in children ( Table 3). The findings were robust in the leave-one-out sensitivity analysis.

Discussion
In the last years, the number of individuals assuming dietary supplements has been steadily increased worldwide [90,91]. Reasons for dietary supplements' use widely varies across the countries: in Europe, it is just limited to general health and well-being, while other countries permit use for medicinal purposes [92].
Considering that dietary supplement production and marketing are usually not strictly subjected to rigid rules as drugs are, there is a need for more data in order to confirm their safe use in the general population and frail subjects.
Pooling data from 71 randomized placebo-controlled clinical studies, this meta-analysis suggests that antioxidant supplementation with ALA was not associated with an increased risk of any treatment-emergent AE. Of note, statistical significance was not even achieved in subsets of studies categorized according to smoking habit, CV disease, presence of diabetes, pregnancy status, neurological disorders, rheumatic affections, renal impairment, and status of children/adolescent.
From a certain point of view, the current analysis strengthens findings from a large observational study considering outcomes data of 610 expectant mothers and their newborns that concluded ALA supplementation is safe in pregnancy even when administered at high doses [93].
These findings are particularly important because they encourage ALA use in a number of conditions in which ALA is actually proven to be effective. As a matter of fact, even though ALA supplementation has already been demonstrated to influence a broad spectrum of metabolic pathways including inflammation and glucose homeostasis [94][95][96], to the best of our knowledge this is the first time that ALA safety profile has been comprehensively evaluated through a pooled analysis of randomized placebo-controlled clinical studies.
Once ALA safety has been established, clinical factors for predicting treatment response should be an objective for future investigations, in order to identify the patient group that might benefit from ALA supplementation the most.
In the past, several meta-analyses showed that ALA supplementation significantly improves both positive neuropathic symptoms and neuropathic deficits to a clinically meaningful degree in diabetic patients with symptomatic polyneuropathy [97][98][99]. Furthermore, ALA was shown to promote weight loss in adults and obese children and adolescents [100,101].
Despite its strengths, this systematic review and meta-analysis has some limitations that mostly inherits from the included clinical studies. First, the effect size on the risk of hypoglycaemic episodes may be affected by variations in the underlying hypoglycaemic therapy in clinical trials enrolling diabetic patients. In fact, the well-recognized euglycaemic effect of ALA may require the adjustment of antidiabetic agents and insulin doses in patients taking antidiabetic drugs [101]. Second, gastrointestinal and CV system AEs included several nosological entities, justifying the probable presence of publication biases for the analysis. However, this limitation is strongly conditioned by the way the AEs were reported in the individual clinical trials. Indeed, most of the studies included in the meta-analysis report the cumulative incidence of gastrointestinal and CV system AEs, without regard to specific type of AEs. Third, AEs were difficult to identify when they were represented by exacerbations of the underlying disease for which ALA was tested (e.g., leg cramps in patients with peripheral polyneuropathy). Moreover, clinical trials testing different ALA regimens often reported the cumulative number of AEs for the supplementation versus placebo. As a result, a sub-analysis by ALA daily dose was not provided. Furthermore, different ALA formulations were tested across the included clinical studies. Despite this, heterogeneity was low for all assessed outcomes, proving that the results were reliable for the whole population and the considered sub-groups [102]. Finally, as per other dietary supplements, a relatively large number of studies have been carried out with open design and/or without a control group, so that they could not be included in a well-carried out meta-analysis.
Future research is needed to understand if sporadic adverse events associated with ALA use are related to the production quality of the used supplements, to other components of mixed supplements and/or to concomitant treatments or diseases, while long-term safety has been already assessed in the NATHAN (Neurological Assessment of Thioctic Acid in Diabetic Neuropathy) 1 trial [84].

Conclusions
Pooling data from the available randomized placebo-controlled clinical studies, the current meta-analysis provides data in support of the safety of the use of ALA to improve health outcomes in overall healthy individuals and in patients affected by other diseases.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-3921/9/10/1011/s1, Figure S1: Plots showing leave-one-out sensitivity analysis for the risk of gastrointestinal AEs following ALA supplementation versus placebo, Figure S2: Funnel plot detailing publication bias for the risk of gastrointestinal AEs following ALA supplementation versus placebo, Figure S3: Plot showing leave-one-out sensitivity analysis for the risk of neurological AEs following ALA supplementation versus placebo, Figure S4: Funnel plot detailing publication bias for the risk of neurological AEs following ALA supplementation versus placebo, Figure S5: Plot showing leave-one-out sensitivity analysis for the risk of psychiatric disorders following ALA supplementation versus placebo, Figure S6: Plot showing leave-one-out sensitivity analysis for the risk of musculoskeletal AEs following ALA supplementation versus placebo, Figure S7: Funnel plot detailing publication bias for the risk of musculoskeletal AEs following ALA supplementation versus placebo, Figure S8: Plot showing leave-one-out sensitivity analysis for the risk of skin AEs following ALA supplementation versus placebo, Figure S9: Funnel plot detailing publication bias for the risk of skin AEs following ALA supplementation versus placebo, Figure S10: Plot showing leave-one-out sensitivity analysis for the risk of infections following ALA supplementation versus placebo, Figure S11: Funnel plot detailing publication bias for the risk of infections following ALA supplementation versus placebo, Figure S12: Plot showing leave-one-out sensitivity analysis for the risk of CV system AEs following ALA supplementation versus placebo, Figure S13: Funnel plot detailing publication bias for the risk of CV system AEs following ALA supplementation versus placebo, Figure S14: Plot showing leave-one-out sensitivity analysis for the risk of hospitalisation following ALA supplementation versus placebo, Figure S15: Plot showing leave-one-out sensitivity analysis for the risk of death following ALA supplementation versus placebo, Figure S16: Funnel plot detailing publication bias for the risk of death following ALA supplementation versus placebo, File A: PRISMA Checklist, File B: Studies excluded from the systematic review after assessment.