This review was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. A comprehensive retrospective search was performed across five major literature databases—PubMed, Web of Science, Cochrane Library, Embase, and Scopus—covering publications up to January 21, 2025. The search strategy of five public databases was provided in Supplementary Table 1. In addition, all relevant systematic reviews containing applicable biomarkers or outcome indicators were manually screened to ensure that no eligible studies were omitted.

Two independent investigators screened the literature and assessed study eligibility based on predefined inclusion and exclusion criteria. In cases of disagreement, a third investigator was consulted to reach a consensus. The initial screening was performed by reviewing titles and abstracts for references to treatment efficacy or biological effects. Studies were included if they met the following criteria:

Studies were excluded if they met any of the following criteria:

Essential data were extracted from each included study, including the first author, year of publication, trial registration ID (if available), assignment model, study population, intervention details (RS type, dose and total intake), control substance characteristics (dose and total intake), intervention duration, and any reported washout period. Participant characteristics were recorded, including the total sample size, allocation to intervention and control groups, sex distribution, mean age, BW, and BMI. Effect sizes and safety outcomes were also collected. Effect sizes were categorized into five domains: (1) Anthropometric parameters, including BW, BMI, waist circumference (WC), hip circumference (HC), waist-to-hip ratio (WHR), fat mass (FM), body fat percentage, diastolic blood pressure (DBP), and systolic blood pressure (SBP); (2) Glycemic profiles, comprising fasting blood glucose (FBG), fasting insulin (FINS), glycated hemoglobin (HbA1c), homeostatic model assessment of insulin resistance (HOMA-IR), and beta-cell function (HOMA-β); (3) Lipid profiles, including TG, total cholesterol (TC), HDL-C, and low-density lipoprotein cholesterol (LDL-C); (4) Inflammatory markers, such as high-sensitivity C-reactive protein (hs-CRP), tumor necrosis factor-alpha (TNF-α), and interleukin-6 (IL-6); (5) Oxidative stress biomarkers, specifically malondialdehyde (MDA) and superoxide dismutase (SOD). Concerning their critical role in functionality, the parameters of intervention substances, including food source, purity, content analytical method, delivery mode, and feasibility, were also reported.

Risk of bias was assessed using the Cochrane Risk of Bias (ROB) tool implemented in Review Manager version 5.3. The tool includes seven domains: (1) random sequence generation, (2) allocation concealment, (3) blinding of participants and personnel, (4) blinding of outcome assessment, (5) incomplete outcome data, (6) selective reporting, and (7) other sources of bias. The overall ROB judgment of each study was provided based on the previous domains. The Jadad scale was used to evaluate methodological quality. The scale awards up to five points based on the following criteria: (1) randomization (1 point for stating randomization, plus 1 additional point for appropriate randomization methods); (2) blinding (1 point for stating double-blinding, plus 1 additional point for appropriate blinding methods); and (3) description of withdrawals and dropouts (1 point for reporting the number and reasons for withdrawals).

Effect sizes were calculated by comparing the mean net changes between the intervention and control groups. The standard deviation (SD) of the net change was estimated using the following formula:

A correction coefficient (R) of 0.5 was assumed for the calculation of standard deviations. For studies that presented data using the median and interquartile range, established mathematical methods were employed to transform these values into the mean and SD (24, 25). If standard error (SE) was reported in some studies, the formula between SD and SE was used to convert to SD using the following formula:

Here, n meant the number of participants in the group. For studies presenting effect sizes in bar charts, mean values and SDs were extracted using ImageJ software. When outcomes were reported separately for subgroups (e.g., males and females), combined means and SDs were calculated using an online statistical tool.¹ Each effect size was expressed as mean difference (MD) with corresponding 95% confidence intervals (CI). In cases where units could not be standardized across studies or when there were substantial differences in measurement scales, standard mean difference (SMD) with 95% CI were calculated. A random-effects model was employed for all meta-analyses to account for inter-study variability. Statistical heterogeneity was assessed using the I² statistic, with values <50% considered low heterogeneity and ≥50% considered high. To explore potential sources of heterogeneity, subgroup analysis was conducted based on RS type (RS2 or RS3), delivery mode (supplement or food), dose (<30 g/day or ≥30 g/day) and duration (<8 weeks or ≥8 weeks) of intervention, geographic region (western developed countries or others), disease type (overweight/obesity, MetS, insulin resistance, prediabetes/T2DM, hyperlipidemia, or NAFLD), mean age (<45 years or ≥45 years), mean BMI (<30 kg/m² or ≥30 kg/m²), overall ROB judgment (low, unclear or high), the score of the Jadad scale (low score of no more than 3 or high score of more than 3) and duration (<8 weeks or ≥8 weeks) of intervention, and assignment (crossover or parallel). Subgroup differences were assessed using Chi-square tests following subgroup meta-analysis. Sensitivity analysis was conducted within studies of non-high ROB judgment to investigate robustness of results. All meta-analyses were performed using Review Manager. Publication bias was assessed visually via funnel plots and statistically using Begg’s and Egger’s tests. If Egger’s test yielded a significant p value for a specific effect size, the Trim and Fill method was applied to estimate adjusted results. Publication bias was performed using Stata version 15.1. A p value < 0.05 was considered statistically significant in the present review.

The study selection process was illustrated in the PRISMA flow diagram (Figure 1). A total of 5,893 records were identified through searches of public literature databases, and an additional 30 records were retrieved from previously published systematic reviews. After removing duplicates, 2,709 unique records remained. Two reviewers (RD and RZ) independently screened the records, and disagreements were resolved by a third reviewer (XM). Of the 2,709 records, 400 articles were deemed potentially eligible based on title and abstract screening. After full-text review, 377 articles were excluded for reasons detailed in Figure 1, resulting in 23 studies included for qualitative synthesis. Among these, 19 studies with extractable effect size data were further included in the quantitative synthesis (meta-analysis).

Table 1 summarized the characteristics of the 23 studies included in this review, published over a 20-year period from 2004 to 2024. Of these, 11 were crossover randomized controlled trials, and 12 were parallel-group trials. Geographically, 11 studies were conducted in Western developed countries: one in Canada (26), one in Denmark (27), one in France (28), three in the United Kingdom (15, 29, 30), and five in the United States (31–35). The remaining 12 studies were conducted in other regions: one in Brazil (36), four in China (10, 11, 37, 38), one in Indonesia (39), three in Iran (40–42), two in South Korea (43, 44), and one in Mexico (45). The included studies covered five categories of metabolic conditions: seven focused on overweight or obesity, six on MetS, eight on T2DM, one on hyperlipidemia, and one on NAFLD. RS doses ranged from 6 to 40 g/day, with intervention durations varying from 4 to 16 weeks. In Maki’s study, the data from higher-dose and lower-dose groups were reported separately; therefore, these two groups were considered as distinct trials in the subsequent analysis. The matching criteria for RS and control interventions were categorized as follows: 12 studies used dose-matched controls, four used carbohydrate-matched controls, five used energy-matched controls, two were unmatched, and one lacked matching details.

A total of 1,073 participants were included across the studies, comprising 430 males and 545 females, with sex-specific data unavailable in two trials. The participants were aged between 31 and 66.1 years, with body weight ranging from 71.8 to 116.1 kg and BMI values between 24.5 and 37.7 kg/m². Effect sizes and safety outcomes are summarized in Table 2. Notably, two studies were identified as having the same registration ID (41, 42). Consequently, to avoid duplication, only one effect size from each study was selected for analysis.

The properties of included studies were documented in Supplementary Table 2. Among these, the most of 20 studies using RS2, and remaining three using RS3. In terms of the sources of RS, 16 studies identified maize as the primary source, two studies focused on banana, one study on potato, one on canna edulis, one on both maize and potato, and one study utilized multiple sources. RS substances had a purity of 34–34.8% when delivered as a powdered supplement in 15 studies, and 3.1–21.2% (26) when delivered as food matrices, including diet, rice, bread, bagel, and snack, in eight studies. The studies utilized Hi-maize 260, whose purity was confirmed by AOAC method 991.43. Besides, two studies employed AOAC 2002.02, two studies utilized the Goni method, one study used a total dietary fiber determination kit, and the remaining three studies did not specify the method used.

Four studies were excluded from the meta-analysis due to the absence of net changes of effect sizes. In Maki’s study, effect sizes for the higher-dose and lower-dose groups were calculated separately. Each effect size included in the meta-analysis was derived from a minimum of three studies. Most of effect sizes were pooled as MDs, while some effect sizes, such as HOMA-β, hs-CRP, TNF-α, SOD, and MDA, were pooled as SMDs.

The results of quality assessment using Cochrane ROB tool were presented in Figure 2 and Supplementary Table 3. All included studies were judged as low risk for random sequence generation. Allocation concealment was assessed as low risk in five studies, while one study was judged as high due to lack of concealment; the remaining 16 studies did not report this information. Blinding of participants and personnel was considered adequate in 20 studies, with three studies rated as high risk in this domain. Regarding blinding of outcome assessment, eight studies were rated as low risk, while 15 studies did not provide sufficient details. For incomplete outcome data, 21 studies demonstrated as low risk, whereas two studies were rated as high due to dropout rates exceeding 20%. Selective reporting and other potential sources of bias were also assessed, and all 23 studies were considered low risk in both categories. For overall ROB judgment, three studies were judged as low risk, 15 were as unclear, and five were as high. The results of the Jadad scale assessments are provided in Supplementary Table 4. In brief, seven studies scored 2 points, two studies scored 3 points, 11 studies scored 4 points, and four studies achieved the maximum score of 5.

Nine anthropometric outcomes were evaluated across the included studies (Figure 3). Nine trials assessed the effects of RS consumption on BW and BMI, with no statistically significant differences observed (BW: MD = −1.33 kg, 95% CI: −3.37 to 0.71, p = 0.20, I² = 96%; BMI: MD = −0.52 kg/m², 95% CI: −1.12 to 0.08, p = 0.09, I² = 92%). In terms of body shape, RS intake was associated with a significant reduction in WC (MD = −2.58 cm, 95% CI: −4.71 to −0.45, p = 0.02, I² = 52%; six trials) and HC (MD = −1.83 cm, 95% CI: −2.03 to −1.64, p < 0.0001, I² = 0%; four trials). The change in waist-to-hip ratio, based on four trials, was minimal and not statistically significant (MD = −0.01, 95% CI: −0.03 to 0.00, p = 0.08, I² = 48%). Regarding fat-related outcomes, RS consumption did not significantly affect FM (MD = −1.55 kg, 95% CI: −3.80 to 0.71, p = 0.18, I² = 97%) or body fat percentage (MD = −1.07, 95% CI: −2.51 to 0.36, p = 0.14, I² = 93%). For blood pressure parameters, no significant effects were observed. SBP showed negligible change (MD = −0.06 mmHg, 95% CI: −3.05 to 2.93, p = 0.97, I² = 69%), while DBP exhibited a non-significant trend toward reduction (MD = −1.47 mmHg, 95% CI: −3.40 to 0.47, p = 0.14, I² = 67%).

Figure 4 illustrated the effects of RS consumption on glycemic profiles. A total of 16 trials evaluated FBG, with no statistically significant effect observed (MD = 0.02 mmol/L, 95% CI: −0.05 to 0.09, p = 0.57), and low heterogeneity (I² = 34%). A meta-analysis of 12 trials demonstrated a significant reduction in FINS levels (MD = −2.39 μU/mL, 95% CI: −3.47 to −1.30, p < 0.0001), although substantial heterogeneity was present (I² = 82%). For HbA1c, a significant reduction was observed with low heterogeneity (MD = −0.14, 95% CI: −0.24 to −0.04, p = 0.008, I² = 3%). HOMA-IR, based on 10 trials, also showed a significant improvement (MD = −0.58, 95% CI: −0.91 to −0.24, p = 0.0008), though heterogeneity was high (I² = 63%). In contrast, three trials reporting on HOMA-β were included in a separate meta-analysis, which showed a non-significant reduction (SMD = −1.08, 95% CI: −2.76 to 0.60, p = 0.21) and high heterogeneity (I² = 92%).

Figure 5 presented the forest plots for the effects of RS consumption on four lipid profile parameters. Across 12 trials, RS intake was not associated with a significant reduction in TG (MD = −0.11 mmol/L, 95% CI: −0.33 to 0.11, p = 0.33), with considerable heterogeneity observed (I² = 88%). However, RS supplementation significantly reduced TC (MD = −0.20 mmol/L, 95% CI: −0.32 to −0.08, p = 0.001, I² = 34%) and LDL-C (MD = −0.11 mmol/L, 95% CI: −0.18 to −0.04, p = 0.003, I² = 12%), both with low heterogeneity. HDL-C levels were not significantly affected by RS intake (MD = −0.02 mmol/L, 95% CI: −0.02 to 0.07, p = 0.29), and substantial heterogeneity was noted (I² = 66%).

Figure 6 illustrated the effects of RS consumption on three commonly reported inflammatory markers. Based on three trials, RS intake did not significantly reduce levels of hs-CRP (SMD = −0.23, 95% CI: −0.52 to 0.07, p = 0.14), with no observed heterogeneity (I² = 0%). For TNF-α, a significant reduction was observed across six trials (SMD = −0.52, 95% CI: −1.00 to −0.05, p = 0.03), although heterogeneity was substantial (I² = 84%). Regarding IL-6, four trials were synthesized, and the pooled analysis revealed no significant effect (MD = −0.12 pg./mL, 95% CI: −0.50 to 0.26, p = 0.54), with high heterogeneity (I² = 65%).

Figure 7 presented the changes in MDA and SOD levels across all subjects. Based on six trials, RS consumption did not significantly reduce MDA levels (SMD = −0.30, 95% CI: −0.65 to 0.05, p = 0.09), with high heterogeneity (I² = 65%). In contrast, data from five trials showed a significant improvement in SOD activity following RS intake (SMD = 0.29, 95% CI: 0.08–0.51, p = 0.008), with no observed heterogeneity (I² = 0%).

Sensitivity analysis indicated that the results were robust in the effect sizes, including HC, TC, LDL-C and SOD. Significant differences yet high heterogeneity persisted in the effect sizes, including WC, FINS, HOMA-IR, and TNF-α. The effect sizes of HDL-C and MDA newly demonstrated a significant enhancement after excluding the studies judges with a high ROB (Supplementary Table 10).

Publication bias was assessed using Begg’s and Egger’s tests, with results summarized in Supplementary Table 11. Begg’s test indicated potential publication bias for HbA1c (P_Begg = 0.027). Egger’s test detected evidence of publication bias for WC (P_Egger = 0.007), DBP (P_Egger = 0.046), FBG (P_Egger = 0.033), and TG (P_Egger < 0.001). Corresponding funnel plots and bias assessments for these five outcomes were presented in Supplementary Figures 1–5. The Trim and Fill methods were applied to WC, DBP, FBG, and TG to adjust for potential bias. For WC, inclusion of three imputed studies yielded a non-significant change in the overall effect size (MD = −4.159 cm, 95% CI: −6.085 to −2.233, p < 0.001). For FBG, the adjusted meta-analysis remained robust, with no additional studies imputed and no change in significance (MD = 0.019 mmol/L, 95% CI: −0.055 to 0.094, p = 0.608). However, significant corrected effect sizes were observed after adjustment in DBP (MD = −2.981 mmHg, 95% CI: −4.825 to −1.136, p = 0.002; 3 imputed studies) and TG (MD = −0.466 mmol/L, 95% CI: −0.657 to −0.276, p < 0.001; 7 imputed studies).

The present review showed RS lowered WC and HC by 2.58 cm and 1.83 cm, respectively. Between the two, WC serves as an indicator of adipose distribution, and abdominal obesity, as determined by WC, is one component of MetS (46). The prior review consistently demonstrated the effect of nondigestible fermentable carbohydrates intake on WC (47). The decrease in WC indirectly indicated a reduction in central adiposity, a pattern of fat distribution, similarly observed in murine models administered with RS (48).

In the current review, neither SBP nor DBP was reduced by RS consumption. However, the findings indicated a reduced impact of RS on DBP following the application of the simulated method. To the best of our knowledge, there is a lack of research investigating the effect of RS on hypertension trials or systematic reviews related to blood pressure. A prior review regarding on the effect of fermentable carbohydrates intake on SBP in adults with overweight and obesity (47). The study on individuals with NAFLD revealed the a significant reduction in SBP and DBP following the consumption of RS for 3 months (11). Another study demonstrated a lessened result of DBP after consuming RS for 12 weeks (35). A possible explanation was that protein intake, compared to carbohydrate intake, was consistently associated with beneficial effect on blood pressure (49).

The meta-analysis showed that consumption of RS improved insulin sensitivity, as indicated by measures such as FINS, HOMA-IR, and HbA1c. Elevated blood glucose levels prompts the secretion of insulin by pancreatic islet beta cells. Insulin resistance is characterized by compensatory hyperinsulinemia. Previous systematic reviews have supported our findings regarding the reduction of FINS level through RS consumption in populations with overweight, obesity, MetS, prediabetes, and T2DM (8, 16, 50). HOMA-IR was used to assess the insulin resistance. Two reviews documented the same results of HOMA-IR (14, 16). The anticipated reduction in FBG levels due to RS intake was not observed. Previous studies demonstrated that RS reduced glucose levels during the postprandial period rather than during fasting (51–53). As for HbA1c, it is conventionally employed as a mean indicator of blood glucose levels to assess glycemic control. We found that HbA1c was lowered by 0.14% after RS consumption. The result was consistent with the preliminary review concerning overweight, obesity, and MetS (8, 16).

The current review identified a reduction in TC and LDL-C concentrations by 0.20 mmol/L and 0.11 mmol/L, respectively, after RS consumption. The effects of RS intake on two biomarkers were consistent across healthy individuals, as well as patients with overweight, obesity and MetS in the previous reviews (16, 47, 54). In a double-blind parallel RCT conducted by Ni et al. (11), it was reported that the intake of RS for 4 months improved TG and HDL-C concentrations. Another study published by Gargari et al. (41) showed that the intake of RS for 8 weeks increased HDL-C levels.

Three inflammatory indicators, including TNF-α, IL-6, and hs-CRP, were reported to be associated with component score of MetS, according to the previous review (55). In the current review, it was demonstrated that the intake of RS led to a reduction in TNF-α levels, while no significant effect was observed on IL-6 levels. These findings are consistent with those reported in previous systematic reviews (16). However, two reviews, which included both healthy subjects and those with diseases and the intervention of RS and resistant dextrin, reported improvements in two biomarkers (56, 57). Previous reviews noted the minimal impact of RS consumption on CRP levels (16–18, 56, 57). Some report indicated hs-CRP is more precise and suitable for detecting elevated cardiometabolic risks than the conventional measurement (58). In the present review, we investigated the impact of RS on hs-CRP levels; however, the finding indicated little effect of RS with low heterogeneity.

As for oxidative stress biomarkers, the concentration of SOD increased, while MDA levels remained unchanged following the consumption of RS. A prior review has provided evidence of improvements in the two biomarkers within the included trials, following interventions with RS and resistant dextrin, in the context of diseases involving chronic kidney disease (17), while another review demonstrated the negative effects of RS intake on the two biomarkers (57). SOD is an enzyme that protects by eliminating superoxide free radicals and their metabolic byproducts. The SOD gene has been reported to be associated with various metabolic disorders (59–61). MDA is produced as a result of lipid peroxidation during metabolic stress. Glucolipotoxicity linked to MetS results in higher levels of plasma MDA (62). Both two reviews reported an enhanced serum level of total antioxidant capacity (TAC), although they exhibited high heterogeneity, including overlapping trials (17, 57). TAC serves as a measure reflecting overall antioxidant buffering power in the body. Conversely, a well-established study indicated that plasma TAC was unable to predict components of cardiometabolic risk, possibly due to the activation of compensatory mechanisms under physiological conditions (63).

Four studies were excluded from the meta-analysis. Of these, three did not provide baseline information, and one concentrated on postprandial changes in serum glucose and insulin. Despite the lack of baseline data, the study by Bodinham et al. (29) demonstrated a reduction in FINS, while Bodinham et al. (30) reported a decrease in TNF-α. Additionally, Robertson et al. (28) observed reductions in FBG, FINS, and HOMA-IR. In contrast, Maziarz et al. (33) found no significant between-group differences in biomarkers.

The classification of RS types was considered to introduce bias, as various in vitro and animal studies have demonstrated that different RS types exhibit variations in digestibility, fermentation rates, and SCFA profiles (21, 64, 65). Only RS2 and RS3 were included in the present meta-analysis, and most (16 of 20) of the trials used RS2. Two records from a well-established study investigating the effect of RS4-enriched flour on metabolic risks were identified, and a significant reduction of TC was observed (66, 67). However, these records were excluded from the current review due to the consumption being ad libitum without a specified dosage, and the inclusion of both healthy individuals and those with MetS. In the present review, subgroup analysis revealed there were no differences between RS2 and RS3 across any parameters. Pugh et al. (68) identified a statistically significant effect of RS2 on FINS level in comparison to RS1. However, the content of RS was not reported in RS1 subgroup, making it difficult determine the differences among RS type. Furthermore, Yuan et al. (54) indicated that RS2, rather than RS3, was associated with the regulation of serum TC and TG, regardless of participants’ health status. We considered there was substantial heterogeneity within the subgroups of two RS types, suggesting that other factors influenced the RS effects.

Food matrix has been considered as a critical contributor to heterogeneity, since starch processing changes not only properties, such as digestibility, gelatinization, retrogradation, crystallinity, amylose to amylopectin ratio, but also bioavailability in recent reports (69–71). Most (13 of 20) of trials encompassed in the current review administered RS in the form of a powdered supplement, thereby preserving the integrity of the heat-labile RS. Conversely, seven studies incorporated RS into food matrices that were subsequently subjected to cooking or baking processes, likely leading to a diminishment in RS content. Moreover, the incorporation of RS flour into food matrices, resulted in a reduced RS content, thereby leading to a decreased actual intake. Thus, the delivery method can affect RS intake and potentially alter outcomes. In the subgroup receiving the supplement delivery, a higher content ranging from 34 to 60% and a dose between 6 g and 40 g were achieved, compared to the subgroup utilizing food matrices, which exhibited a content range of 3.1–21.2% and a dose of 4–30.9 g. Most of RS-rich powders using in the included studies were commercially manufactured Hi-maize 260. In the subgroup receiving RS as a supplement, body fat percentage, SBP, and DBP significantly decreased, but these effects were absent when RS was part of food matrices. Wei et al. (57) reported the supplement form influenced the RS effects on inflammation biomarkers. Unfortunately, due to the limited information provided in the included studies, details regarding the properties of intervention, such as amylose to amylopectin ratio or crystallinity, were not available within the context of the articles.

Subgroup analysis on the dose and the duration of intervention was conducted. The result showed that consuming 30 g/day of RS improved anthropometric parameters, such as BW, WC, body fat percentage, SBP and DBP, significantly. Johnston et al. (15) reported that the RS supplement was integrated into the daily food twice per day. Li et al. (12) proposed that the RS supplement be administered as a powder dissolved in 300 mL of water. Meanwhile, Maziarz et al. (33) ensured an adequate RS content by incorporating RS powder into muffins. Two reviews set the dose cut-off at 20 g/day (18, 54), but only one reported its role of the mediator in RS effect on TC level. We considered the limited role of the dosage appeared in previous reviews, as it did not meet the recommended dietary fiber intake (72). As for the duration, according to published reviews, RS has been demonstrated to reduce serum inflammation biomarkers for a loner duration (56, 57). Here, we found the subgroup with over 8 weeks of intervention showed greater reductions in body fat percentage and DBP compared to those with less than 8 weeks.

We conducted subgroup analysis on participants characteristics, such as region, disease, mean age and mean BMI. The world’s division into western and eastern regions influenced the RS effects on glycemic, lipid and inflammatory profiles, as noted in two previous reviews (16, 57). We found that the RS consumption significantly decreased the levels of FM and MDA in the participants from outside western developed region. The RS effects in the current review were partly influenced by the disease profile. The result was cautiously presented, as only single trial with disease, such as hyperlipidemia, or NAFLD, was included in the review. In the younger subgroup, RS consumption significantly affected BW, FM, body fat percentage, and HDL-C, while it reduced MDA level in the middle and elder subgroup. Previous reviews also noted significant RS effects on glycemic, lipid, and inflammatory biomarkers in the younger subgroup (16, 57). RS consumption led to reductions in BW, FM and body fat percentage was decreased among overweight participants. We concerned RS consumption, as the supplement intervention, exhibited greater sensitivity in younger individuals and those with a lower BMI.

Subgroup analysis on study design based on Cochrane ROB tool, the score of the Jadad scale, and assignment, was explored as well. The levels of TNF-α and MDA were significantly improved by RS intake in the subgroups of trials judged as low or unclear ROB, while the parameters of BMI, body fat percentage, HOMA-IR, HDL-C, and MDA were enhanced in the subgroups of trials achieved a higher score of the Jadad scale. Both quality assessment tools were employed in systematic reviews. The first tool was evaluated across a broad range of aspects, whereas the second scale emphasized rigorous randomization and blinding methods. The results of the subgroup analysis indicated the necessity for a high-quality study design to research on the clinical functionalities of RS. Differences of assignment was only influenced the results of MDA. RS intervention is highly susceptible to carryover effects due to microbiota adaptation. Therefore, we recommend that a longer washout period be implemented in crossover trials.

In relation to mechanisms of regulating anthropometric and serum biomarkers, the functional benefits of RS consumption can be attributed to either its physical–chemical structure or its biological impact.

Furthermore, the application of RS enhances the water absorption capacity of doughs and diminishes the viscoelastic properties of foods (73), thereby potentially augmenting satiety and decreasing the intake of digestible carbohydrates. RS-containing diets notably decreased energy intake and BW in vivo (74). In human, the preliminary review gave inconclusive evidence that RS had effects on appetite, hunger, food intake, satiety (75). The malabsorption of RS in the small intestine leads to a reduction in the glycemic index following ingestion, resulting in minimal fluctuations in postprandial blood glucose levels, which effectively enhances glucose tolerance (76). RS3 can be fermented in the colon and may bind bile salts, reducing bile acid reabsorption in the ileum, stimulating hepatic bile acid production, and increasing cholesterol use (77).

The mechanism of RS consumption may be partially attributed to the indirect effects mediated by the gut microbiome and the production of SCFAs. Indigestible RS are fermented by the predominant beneficial gut microbiota, such as Bifidobacterium, Lactobacilli, Ruminococcus, Roseburia, Eubacterium, Faecalibacterium, and Clostridium among others (78–80). Bifidobacterium, Faecalibacterium and Ruminococcus were typically increased after RS2, RS3, and RS4 supplementations, respectively, according to published trials (10, 37, 67). Fermentation of RS generates SCFAs including acetate, propionate, butyrate (81). RS can reduce appetite by enhancing the secretion of satiety-associated hormones like glucagon-like peptide-1 (GLP-1) and peptide YY (PYY) by SCFAs targeting G protein-coupled receptors (GPRs) (82). It was found that muffins with RS content enhanced satiety, prolonged digestion period, and benefited to weight loss (33). Replacing carbohydrates with RS led to a notable increase in postprandial fat oxidation (83). Additionally, the fermented SCFAs shifted metabolic pathway from lipogenesis to fat oxidation by activating peroxisome proliferator-activated receptor-γ (84). These findings suggest that RS may contribute to reducing fat accumulation.

Fermented SCFAs could reduce the gastric motility and intestinal glucose transport capacity by stimulating PYY and GLP-1, respectively (85, 86). The Dose-dependent Kudzu RS enhanced insulin sensitivity and reinstated the hepatic expression of insulin receptor substrate 1 and glucose transporter 4, both of which contributed to insulin efficacy and glucose homeostasis, in T2DM mice (87). Serum SCFAs have been shown to improve insulin sensitivity by interacting with G protein-coupled receptors (GPRs) [61]. Butyrate has been shown to mitigate adipocyte inflammation by inhibiting the NOD-like receptor thermal protein domain associated protein 3 pathway [71]. Bacterial toxins, particularly lipopolysaccharides (LPS), which are integral components of Gram-negative bacteria, are primary contributors to the cause of low-grade chronic systemic inflammation (88). Gut microbiota-derived LPS affect intestinal permeability and tight junction (TJ) proteins by activating toll-like receptor 4-dependent pathways in intestine (89). Compromised intestinal permeability led to LPS translocation into the blood. Dietary RS addition increased propionate and butyrate and ameliorated LPS-induced inflammation during LPS challenge in vivo (90). Both SCFAs improve the intestinal barrier function by upregulating TJ proteins in epithelial intestine cells (91, 92). LPS induces inflammatory cytokines secretion, like TNF-α, by mediating inflammatory host cells, such as macrophage (93). Chronic inflammation induces lipid peroxidation through inhibition of catalase and glutamine peroxidase activities (94). SCFAs play protective role in inflammatory cytokines and oxidative stress (94, 95). However, there have been no reports demonstrating that SCFAs fermented by RS modulate inflammation and oxidative stress. The probable mechanism is that the fermentation of RS reduces the diversity and expression of hazard gut microbiota (7).

Dietary RS mediates bile acid (BA) metabolism and promotes GLP-1 secretion by targeting Takeda GPR 5 (96). Lei et al. demonstrated lotus seed RS lowered serum lipid by promoting BAs excretion and regulating gut microbiome (97). Included trials reported distinct changes of BA metabolism after RS2 and RS3 (10–12).

We implemented strict eligibility criteria for including RCTs, standardized algorithms and uniformity in the units before calculating effect sizes. The heterogeneity was evaluated from the variety of aspects, such as RS type, intervention settings, participant characteristics, and study design. We employed a variety of measurement tools for qualitive assessments, sensitivity analysis and publication bias.

However, our study still had several limitations. Additional detailed parameters of RS, such as crystallinity, amylose to amylopectin ratio, gelatinization, and retrogradation, were not collected, even though these factors potentially influenced the observed effects. It was hard to draw a solid conclusion due to the absence of proposed mechanisms in the majority of the included studies. The present review focused on the limited scope of adults excluding children or adolescents. We included the participants with metabolic risks, including overweight, obesity, MetS, prediabetes, T2DM, hyperlipidemia, and NAFLD. While these conditions exhibited similarities in terms of pathogenesis and symptoms, it was important to acknowledge the distinctions that existed between them.