The Emerging Role of Flavonoids in Autism Spectrum Disorder: A Systematic Review

Although autism spectrum disorder (ASD) is a multifaceted neurodevelopmental syndrome, accumulating evidence indicates that oxidative stress and inflammation are common features of ASD. Flavonoids, one of the largest and best-investigated classes of plant-derived compounds, are known to exert antioxidant, anti-inflammatory, and neuroprotective effects. This review used a systematic search process to assess the available evidence on the effect of flavonoids on ASD. A comprehensive literature search was carried out in PubMed, Scopus, and Web of Science databases following the PRISMA guidelines. A total of 17 preclinical studies and 4 clinical investigations met our inclusion criteria and were included in the final review. Most findings from animal studies suggest that treatment with flavonoids improves oxidative stress parameters, reduces inflammatory mediators, and promotes pro-neurogenic effects. These studies also showed that flavonoids ameliorate the core symptoms of ASD, such as social deficits, repetitive behavior, learning and memory impairments, and motor coordination. However, there are no randomized placebo-controlled trials that support the clinical efficacy of flavonoids in ASD. We only found open-label studies and case reports/series, using only two flavonoids such as luteolin and quercetin. These preliminary clinical studies indicate that flavonoid administration may improve specific behavioral symptoms of ASD. Overall, this review is the first one to systematically report evidence for the putative beneficial effects of flavonoids on features of ASD. These promising preliminary results may provide the rationale for future randomized controlled trials aimed at confirming these outcomes.


Introduction
Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder characterized by deficits in social interaction, impairments in language and communication abilities, and repetitive/stereotyped behaviors [1]. Emerging research suggests that the global prevalence of ASD has increased considerably over time, indicating a median prevalence of 100/10,000 within and across regions [2]. The vast majority of individuals with ASD do not receive an etiological diagnosis and receive a diagnosis of autism of unknown etiology. At this time, even after a thorough evaluation, the majority of cases of ASD have an unknown cause. Although there is no consensus on the causes of ASD, multiple risk factors have been proposed. Currently, the relationship between genetics and environment is thought to be a key driving force in the pathophysiology underlying ASD [3]. Large-scale genetic studies have identified hundreds of genes that have a role in synaptic development and function as risk factors for ASD pathogenesis. Genetic alterations associated with ASD may include different classes of genetic variants, such as single nucleotide polymorphisms effects may contribute to alleviating the pro-inflammatory state of children with ASD that exhibit heightened stress reactivity and hyperarousal symptoms. Furthermore, flavonoids may interact with a wide variety of neuronal signaling cascades, enhancing neuro-cognitive performance and increasing neurogenesis under healthy or pathological conditions [26]. Some natural flavonoids may also exert anxiolytic action through the activation of benzodiazepine receptors [27]. So far, the effect of flavonoids on ASD has not been systematically analyzed. Therefore, we aimed to systematically review all available findings generated from both preclinical and clinical studies investigating the role of flavonoids on ASD.

Methods
This study followed the Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) guidelines for systematic reviews [28].

Eligibility Criteria and Data Extraction
We included preclinical and clinical studies that investigated the effect of flavonoid interventions on outcomes associated with ASD. Only reports that estimated the flavonoid content of foods or dietary supplements were included. We excluded articles for the following reasons: articles such as reviews, meta-analyses, conference papers, and book chapters and studies not published in English. The titles and abstracts obtained from the databases were independently reviewed by two authors (R.S. and S.A.). The full-text screening was conducted, excluding studies that did not meet the inclusion criteria. A third author (S.D.) was consulted in the case of disagreement about the eligibility of a study. In cases where full text was not available, we contacted the corresponding author and asked him to provide full-text publications within a 1-week time frame. The authors developed a data extraction form on an Excel sheet and the following data from eligible studies were extracted: author's name; publication year; experimental model; study design; subject characteristics; intervention (duration, type of compounds, and dose); and results.

Risk of Bias
The risk of bias in the included preclinical studies was evaluated using the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) tool [29]. This tool was developed to assess methodological quality and measure the bias in studies involving animal models. The SYRCLE tool considers the following domains: sequence generation, baseline characteristics, allocation concealment, random housing, investigator blinding, random outcome assessment, outcome assessor blinding, incomplete outcome data, and selective outcome reporting. These domains are related to five types of bias: selection bias; performance bias; detection bias; attrition bias; and reporting bias. For each included study, the bias types were classified as "high", "low", or "unclear". Data on the housing conditions, such as light/dark cycle and temperature, were also extracted as an additional indicator of study quality.

Selected Studies
As shown in Figure 1, we retrieved a total of 383 published studies from the three databases, but 178 were duplicates. We discarded 141 articles during the screening because they did not meet the inclusion criteria. We examined the remaining 64 articles for eligibility through full-text reading. Of these, 43 studies did not meet our inclusion criteria, or the full text was unavailable. Thus, a total of 21 studies were included in the final qualitative analysis. Seventeen papers specifically addressed the effects of flavonoids in animal models of ASD, 4 on ASD patients, and 1 study used both mice and humans as experimental systems. We present the main results of these studies in the following sections. random outcome assessment, outcome assessor blinding, incomplete outcome data, selective outcome reporting. These domains are related to five types of bias: selection performance bias; detection bias; attrition bias; and reporting bias. For each inclu study, the bias types were classified as "high", "low", or "unclear". Data on the hou conditions, such as light/dark cycle and temperature, were also extracted as an additi indicator of study quality.

Selected Studies
As shown in Figure 1, we retrieved a total of 383 published studies from the t databases, but 178 were duplicates. We discarded 141 articles during the screening cause they did not meet the inclusion criteria. We examined the remaining 64 article eligibility through full-text reading. Of these, 43 studies did not meet our inclusion c ria, or the full text was unavailable. Thus, a total of 21 studies were included in the qualitative analysis. Seventeen papers specifically addressed the effects of flavonoid animal models of ASD, 4 on ASD patients, and 1 study used both mice and human experimental systems. We present the main results of these studies in the following tions.

Preclinical Studies
The study characteristics and outcomes of the preclinical studies are summarized in Table 1. An extract of Bacopa monnieri, a nootropic herb, has been used to evaluate its neuroprotective effect in a valproic acid (VPA) model of ASD. The most abundant compound identified in this extract was luteolin, followed by apigenin. These flavonoids belong to the subclass of flavones. The results showed that B. monnieri extract, administered postnatally to rat pups at 80 mg/kg, may attenuate VPA-induced damage by restoring antioxidant enzymes and reducing inflammatory cytokines in the hippocampus and prefrontal cortex. In these brain regions, the treatment also reduced mRNA and protein expression of AMPA receptor, which plays a vital role in neurodevelopmental disorders such as ASD [30]. Furthermore, the in-silico analysis also showed a good binding profile of luteolin against the competitive antagonist binding site on the AMPA receptor. These effects were accompanied by improvements in learning and memory impairments, repetitive behavior, motor coordination, and social deficits [31].
Using a VPA model of autistic behaviors, the association of luteolin with palmitoylethanolamine (PEA) ameliorated autistic-like behavioral changes, including reduced sociability and increased anxiety-related behavior. The treatment not only reduced the expression of proinflammatory mediators, such as NF-kB, interleukin-1 beta (IL-1β), and tumor necrosis factor-alpha (TNF-α) but modulated apoptosis markers (Bax and Bcl-2) in hippocampus and cerebellum, also increasing neuroplasticity and neurogenesis in the hippocampus [32].
Naringenin, a flavonoid belonging to the flavanone subclass, has been reported to restore behavioral and biochemical deficits in a 3-4 months old male propanoic acid (PPA) rat model of ASD. Treatment was started on the 2nd day post-surgery and was continued till the 29th day. After the treatment period, naringenin encapsulated in polylactic-co-glycolic acid (PLGA) nanoparticles reduced the expression of matrix metalloproteinases-9 (MMP-9) and heat-shock proteins 70 (HSP-70). These proteins may have functions in driving the neuroinflammatory state associated with ASD [33]. In the same model, naringenin improved mitochondrial function in the brain by restoring, at least in part, the activities of mitochondrial enzyme complex I and II. Furthermore, naringenin nanoparticles reduced the expression of P-glycoprotein (P-gp), which is a transporter responsible for preventing the entry of various therapeutic moieties across the blood-brain barrier (BBB). These effects were also accompanied by improvements in sociability and perseverative behavior [33].
Quercetin, ubiquitous in plant-based foods and beverages, is categorized as flavanol. Using a VPA rat model, quercetin administered over 13 days (from the 6th to the 28th day of gestation) prevented alterations in social interaction and nociception in the rat pups. Likewise, treatment with quercetin prevented brain damage by improving oxidative stress parameters, mainly in the hippocampus and striatum [34].
Baicalin is a flavonoid of high biomedical value isolated from the root of Scutellaria baicalensis. Elesawy et al. demonstrated that postnatal treatment with baicalin might ameliorate neurochemical and behavioral alterations in a VPA rodent model of ASD. Specifically, baicalin improved neuronal mitochondrial functions, as demonstrated by increased synthesis of mitochondrial adenosine triphosphate (ATP) level and enhanced expression of mitofusin-2. This flavonoid elevated the level of sirtuin-1 (SIRT1) in the brain tissues and restored antioxidant enzymes such as superoxide dismutase (SOD) and catalase (CAT). Improvements in motor development, repetitive behavior, and social deficits have also been observed [35].  One study used a genetic loss-of-function model of an ASD risk gene (CNTNAP2) to conduct a pharmacological screen and identify novel compounds against ASD. The authors found that biochanin A, a phytoestrogenic isoflavone, might reverse the mutant behavioral phenotype in zebrafish larvae by decreasing night-time activity. Although biochanin A activated the expression of estrogen response genes, this transcriptional activation appeared to be independent of the behavioral rescue. In addition, early exposure to biochanin A did not reverse the GABAergic deficits in this model [36].
The 7,8-dihydroxyflavone (7,8-DHF) is a flavonoid that mimics the physiological actions of brain-derived neurotrophic factor (BDNF), activating tyrosine receptor kinase B (TrkB) and promoting neuronal survival, synpatogenesis, and axonal regeneration. Johnson et al. demonstrated that 7,8-DHF may attenuate some ASD symptoms in a Rett syndrome (RTT) mouse model. Oral administration of 7,8-DHF throughout life extended lifespan, increased the size of neuronal nuclei, and enhanced voluntary locomotor activity in these mice. In addition, 7,8-DHF partially ameliorated irregular breathing patterns and restored tidal volumes to wild-type levels [37]. In another study, 7,8-DHF reversed the altered synaptic structure and function caused by the genetic deletion of vaccinia-related kinase 3 (VRK3). VRK3 plays essential roles in synaptic structure and cognitive functions through the regulation of extracellular signal-regulated kinase (ERK), which is involved in the regulation of synaptic protein synthesis, dendritic morphology, and functional plasticity. Moreover, TrkB activation by 7,8-DHF treatment restored social interactions in VRK3deficient mice showing autism-like behavior (12-15 weeks old) [38].
Genistein is a xenoestrogen (isoflavone) that may interfere with the development of estrogen-sensitive neural circuits and detrimentally affect the offspring microbiome-gutbrain axis. Many fetuses and infants are exposed to xenoestrogens through the placenta and milk. This exposure may increase the risk for ASD. One study exposed California mice offspring to bisphenol A (BPA), showing that these animals spent more time engaging in repetitive behaviors, which is considered a type of autistic-like behavior. Similarly, mice (90 days of age) exposed to genistein through the maternal diet engaged in similar repetitive behaviors and showed socio-communicative disturbances. These effects may be due to alterations in carbohydrate metabolism, phenylalanine, and tyrosine metabolism [39].
A study conducted by Khalaj et al. demonstrated that prenatal treatment with hesperetin, a flavonoid belonging to the flavanone subclass, may ameliorate autistic-like behaviors and oxi-inflammatory parameters (e.g., SOD, CAT, IL-6, and TNF-α) in the brain of rat pups exposed to VPA. Likewise, histopathological findings indicated that hesperetin protected Purkinje cells of the cerebellum [40]. Treatment with the flavanol catechin, a flavonoid derived from green tea, has been proven to target the nitric oxide pathway and ameliorate behavioral, biochemical, neurological, and molecular deficits in a PPA rat (3-4 months) model of ASD. Moreover, this compound improved the levels of neuroinflammatory and apoptotic markers such as TNF-α, IL-6, NF-κB, interferon-gamma (IFN-γ), HSP-70, and caspase-3 [41].
Alpha-glycosyl isoquercetin (AGIQ) is a flavonol glycoside that can be found in citrus fruits, red beans, and buckwheat. Continuous AGIQ treatment, starting during late gestation, ameliorated lipopolysaccharides (LPS)-induced pro-inflammatory responses and oxidative brain damage during infancy and prevented the expression of subsequent deficits in neurogenesis and behavior throughout the adult stage [42]. In another study, epigallocatechin gallate (EGCG), the most prevalent flavanol of green tea, alleviated neurological damage in a PPA 21-day-old rat model of ASD by increasing nerve growth factor (NGF), BDNF, TrkB, and calcium/calmodulin-dependent protein kinase II subunit alpha (CaMKII-α) levels and decreasing cAMP response element-binding protein (CREB) levels [43].
Prenatal prophylaxis with diosmin, a flavonoid structural analog of luteolin, showed to inhibit neuronal JAK2/STAT3 phosphorylation following the IL-6 challenge in a maternal immune activation (MIA) model of ASD. This flavone also improved behavioral deficits in social interaction in adult offspring [21].
Anthocyanins, a class of flavonoids present in berry fruits, are considered promising agents to reduce microglia-driven neuroinflammation. Serra et al. demonstrated that an anthocyanin-rich extract alleviated autism-like behaviors in a VPA-mouse model. At the same time, this extract decreased both neuroinflammation and gut inflammation, modulating the composition of the gut microbiota. Increased levels of serotonin and reduced synaptic dysfunction have also been demonstrated [44].
The Cdkl5 knockout (KO) mouse model is characterized by ASD features, intellectual disability, and early-onset epilepsy. The chronic administration of luteolin ameliorated hyperactive profile, memory ability, and motor stereotypies in this model. Moreover, this flavonoid also improved dendritic spine maturation and dendritic arborization of cortical neurons, increasing hippocampal neurogenesis [45]. In the same model, defective synaptic maturation in the hippocampi and cortices can be rescued through the intraperitoneal administration of EGCG, which is, however, not sufficient to normalize behavioral CDKL5dependent deficits. Green tea flavonoid EGCG also restored defects in dendritic and synaptic development of primary Cdkl5 KO neurons [46].

Risk of Bias in Preclinical Studies
As shown in Table 2, most included preclinical studies (16 out of 17) described baseline characteristics. Nine studies showed a high risk of bias due to a lack of information on the technique for sequence generation, while seven studies were determined as an unclear risk since this information was not clearly stated. The allocation concealment was not described in detail in most of the studies (16 out of 17). An unclear risk of bias was determined for all studies due to ambiguity with random housing, but all studies clearly described housing conditions (e.g., light/dark cycle, temperature, and humidity). Data regarding investigator blinding was unclear in 11 studies. It was also unclear in most of the studies (14 out of 17) whether animals were selected at random for the outcome assessment. Blinding outcome assessment was reported in eight studies, while there was no clear evidence of blinding of the outcomes assessor in nine studies. The reporting of incomplete outcome data was unclear in 12 studies, while 4 studies included data collection from all outcome results. High risk was determined for one study due to missing outcome data. Sixteen studies were considered at low risk of bias from selective outcome reporting.

Clinical Studies
Five clinical studies involving a total of 145 children investigated the effects of flavonoids on ASD featuring (Table 3). However, one of these studies has already been presented in Section 3.2 since it used mice and humans as experimental systems. Similar to the results obtained from the mouse model, Bertolino et al. reported that a combined treatment of luteolin and PEA for 12 months improved the clinical picture in a 10-year-old male child with a reduction in stereotyped behaviors [32].
A 26-week, prospective, open-label trial demonstrated that a dietary supplement formulation containing luteolin and quercetin might provide significant benefits in ASD children both in adaptive functioning and behavioral difficulties. These flavonoids are considered safe, and the only adverse effect noted in the subjects was transient irritability [48]. Using the same flavonoids at the same dose, an uncontrolled open case series showed that treatment with luteolin and quercetin for 4 months might increase attention and sociability in children with ASD. The authors also reported an improvement in gastrointestinal dysfunction that may have had a substantial impact on the improvements seen in these children [49]. Likewise, an open-label trial on a cohort of 40 ASD children showed that the serum levels of IL-6 and TNF decreased significantly after a treatment period of 26 weeks with luteolin and quercetin, as compared with normotypic controls. This study also indicated a positive effect of luteolin and quercetin on the adaptive functioning of this cohort of ASD children [50].

Discussion
The scope of the present review was to systematically synthesize the current preclinical and clinical evidence on the effects of flavonoids in ASD and its associated symptoms. The majority of included studies found a positive result, suggesting that flavonoid administration may improve ASD features, including impairment in socialization and repetitive and stereotypic behaviors. Although all the included clinical studies found that flavonoids may help to mitigate the behavioral issues of ASD, there is a general paucity of randomized placebo-controlled trials evaluating the use of these compounds in children with ASD. Although placebo-controlled trials are considered the "gold standard" in medical research, the use of placebos in the pediatric field poses ethical and scientific challenges. First, children cannot exercise the principle of autonomy and are subject to parental decision-making on their behalf. Second, the placebo response might be larger in children/adolescents than in adults [51,52]. A recent study meta-analyzed the placebo response of core symptoms in pharmacological and dietary supplement ASD trials. Although no difference was found between age groups, these results should be interpreted with caution because the majority of studies were in pediatric populations. In order to increase the detection of the efficacy of experimental interventions for ASD, the same study also suggests considering the predictors of placebo response, such as the use of a threshold of core symptoms at inclusion, caregiver ratings, and flexible dosing. When large sample sizes and multiple sites are required, they should be carefully selected, trained, and monitored, trying to keep the number of sites at the minimum feasible [53]. In this review, examining only open-label studies and case reports/series, there is not sufficient evidence to support the clinical efficacy of flavonoids in ASD patients. Moreover, all the clinical studies had small sample sizes and used only two flavonoids such as luteolin and quercetin.
Luteolin is a common flavonoid present in many fruits, vegetables, and herbs. Although its bioavailability is low, luteolin represents one of the most powerful and effective flavonoids, which has displayed numerous biological properties, including antiinflammatory, antioxidant, and neuroprotective properties [54]. Several studies indicate that the anti-inflammatory and antioxidant effects of luteolin are mediated through the inhibition of NF-kB and induction of redox-sensitive transcription factors involved in the activation of antioxidant defense systems [55,56]. Luteolin is also structurally related to 7,8-DHF, which was shown to have BDNF-like activity. In fact, it has been shown that luteolin may induce hippocampal neurogenesis by promoting the activation of BDNF [57]. A case report included in our review used luteolin in combination with PEA, an endocannabinoidlike lipid mediator with lipophilic nature. Many studies demonstrated that combined treatment with these compounds might stimulate both hippocampal neurogenesis and dendritic spine maturation to a greater extent than either luteolin or PEA alone [58,59]. Luteolin and quercetin share structural chemical features, and similar findings have been reported using quercetin. Despite some controversial results, quercetin increases survival against oxidative insults, providing neuroprotection through modulation of transcription factors and survival signaling cascades associated with antioxidant and anti-inflammatory pathways [60,61]. However, although delivery strategies are being developed (i.e., nanoformulations and lipid carriers), absorption and metabolic studies showed that quercetin and luteolin have very limited bioavailability [62,63]. Overall, the clinical utility of flavonoids, including luteolin and quercetin, to manage behavioral symptoms in patients with ASD remains to be validated by future clinical studies.
Due to the limited availability of postmortem brain tissue to determine the cellular and molecular alterations associated with ASD, animal models may help to investigate the neural structure of the autistic brain and define the neural systems that constitute the social brain and mediate repetitive behaviors. These preclinical models can also be employed to test the safety and effectiveness of potential therapeutic compounds [64]. Although animal models may have great translational value, the limitations of ASD models are rarely acknowledged, and the predictive validity of these models for humans is often overstated or misinterpreted. However, a recent literature review provided recommendations to identify limitations such as minimum sample sizes, sex controls, breeding schemas, housing conditions, genetic background, and task validation [65]. Despite this, a considerable number (16 out of 17) of the preclinical studies included in this review observed significant results of flavonoids against neurobehavioral alterations associated with ASD. Only one article reported a negative result, showing that genistein, a soy-derived isoflavone, may induce repetitive behavior and promote socio-communicative disturbances in California mice. These effects might be due to alterations in the microbiome-gut-brain axis induced by genistein exposure [39]. However, discordant results have been reported on the effects of genistein in animal behavior studies. Perinatal exposure of rats to genistein improved spatial learning and memory but impaired passive avoidance learning and memory [66]. Other studies with adult rodent models further suggest that genistein improves spatial and placement learning and memory [67,68]. Conversely, another study showed that male rats exposed to genistein through the maternal diet during both gestation and lactation exhibited spatial learning and memory deficits [69]. In humans, a recent national population-based observational cohort study examined the long-term neurodevelopmental outcomes during childhood following the consumption of soy formula rich in isoflavones during infancy.
There was no evidence that soy formula increases the risk of epilepsy and ASD [70].
The preclinical studies included here used different animal models to investigate the effect of flavonoids on the behavioral and neurochemical characteristics of ASD. The VPA rodent model was the most frequently used model of ASD, followed by other experimental systems, such as the PPA rat model and genetic models. A good variety of flavonoids and their representative subclasses has been investigated in the research reports included in this review. One of the most examined flavonoid subclasses in the preclinical studies was flavones with a series of compounds such as luteolin, apigenin, baicalin, 7,8-DHF, and diosmin. Other flavonoid subclasses were flavanones (naringenin and hesperitin), flavonols (quercetin), isoflavones (genistein and biochanin A), flavanols (catechin and EGCG), and anthocyanins.
ASD neurobiology is thought to be associated with oxidative stress, as shown by increased levels of reactive oxygen and nitrogen species and alterations in other indicators of oxidative stress. ASD is also characterized by decreased glutathione reserve capacity. In particular, low levels of reduced glutathione (GSH), high levels of oxidized glutathione (GSSG), and alterations in the expressions of glutathione-related enzymes in the blood or brain appear to be important factors in the pathogenesis of ASD [9,71]. Findings from preclinical studies suggest that treatment with flavonoids, such as luteolin, quercetin, hesperetin, and catechin, can increase the activity of antioxidant enzymes, such as SOD, CAT, glutathione reductase (GRx), and peroxidase (GPx). Similarly, flavonoids may also reduce ROS and nitrite levels as well as decrease malondialdehyde MDA levels in ASD experimental models [31,34,[40][41][42].
In recent years, the contribution of inflammation and neuro-immune dysregulations to ASD has been the object of intense research. Several studies have repeatedly found that increased systemic levels of pro-inflammatory mediators, altered patterns of immune cell response to activation stimuli, and abnormal microglia activation are hallmarks of ASD [72,73]. Consistent with their anti-inflammatory action, preclinical findings suggest that flavonoids, especially naringenin and hesperetin, can downregulate the expression of inflammatory mediators, such as IL-1β, IL-6, and TNF-α, through inhibition of NF-kB [31][32][33]40]. Isoquercitrin and anthocyanin may also reduce the reactivity of microglial activation markers, such as Iba1, CD68, and CD11, as well as astrocyte marker glial fibrillary acidic protein (GFAP) [42,44]. Some experimental studies also revealed that these effects are accompanied by improvement of mitochondrial function and gut microbiota composition. Interestingly, another possible mechanism, only partially explored in this range of studies, is that flavonoids, such as baicalin, naringenin, and anthocyanin, may modulate mitochondrial ATP production and factors involved in the respiratory chain deficiency [33,35,39,44]. Alongside an amelioration of the behavioral phenotype of ASD, findings from animal models seem to suggest a pro-neurogenic effect of flavonoids. Luteolin, EGCG, and isoquercitrin can activate the expression of neurotrophic factors, such as BDNF and nerve growth factor (NGF) [42,43,45]. The risk of bias in animal studies was evaluated according to the SYRCLE tool. The methodological quality of many studies was unclear since many items were unclear due to a lack of precise information. This suggests that there is much room for improvement. However, it is important to mention that the items in the SYRCLE tool are quite difficult to assess in animal intervention studies at present because protocols for animal studies are not yet registered in central, publicly accessible databases [29].
It is also important to note that the differences in bioavailability and absorption rates of the flavonoids are lacking in the included preclinical and clinical studies. The structure of flavonoids influences the rate of intestinal absorption, and several studies also suggest that the metabolites of flavonoids may be one of the characteristics responsible for their beneficial effects [74]. Therefore, given that flavonoid metabolism exhibits extensive variation between individuals, these aspects should be considered in future studies. Moreover, most of the preclinical studies examined here used different treatment durations and a large variety of flavonoid dosages. The dosages of flavonoids in animal studies are usually higher than those that are achievable by usual dietary intakes in humans.
The strength of this article is that it is the first review to systematically report evidence for the putative beneficial effects of flavonoids on ASD. Another strength of the current review is its broad scope and comprehensive search strategy, as we wanted to include a wide range of flavonoids and evaluate their effects on ASD features. However, this review has limitations which need to be acknowledged. Given the large heterogeneity among studies, no cumulative meta-analysis was conducted. This was due to the high heterogeneity among the studies with several experimental models used, distinct methodologies in the analyzed parameters, and high variation related to the dose and duration time of flavonoid supplementation. Another limitation of our review is that the risk of bias in the clinical investigations was not conducted due to the small number of the included studies and the lack of a reliable and adaptable tool to evaluate the methodological quality of these studies. However, the most important limitation of the present review is the lack of randomized, placebo-controlled trials, which prevents the strength of the conclusions that can be drawn from this review.

Conclusions
This systematic review summarizes the potential benefits of flavonoids on molecular and behavioral aspects of ASD. However, despite a variety of preclinical investigations that have been conducted, supporting, at least in part, the putative beneficial effects of flavonoids against ASD, blinded randomized clinical trials are needed. Thus, large-scale and well-designed controlled trials are essential to validate the preclinical findings and identify the most effective strategy (type of flavonoids, concentration, treatment duration) for patients with ASD.
Author Contributions: R.S., M.M. and S.D.: conceptualization and project development; R.S., A.M., S.A. and S.D.: data collection, data extraction, manuscript drafting; R.S., A.M., G.S., M.M. and S.D.: manuscript writing, review, and editing. All authors contributed to this article and approved the submitted version. All authors have read and agreed to the published version of the manuscript.