Assessing the effects of inulin‐type fructan intake on body weight, blood glucose, and lipid profile: A systematic review and meta‐analysis of randomized controlled trials

Abstract Inulin‐type fructan (ITF) intake has been suggested to alleviate several features of metabolic syndrome including obesity, diabetes, and hyperlipidemia; yet, results from the human trials remained inconsistent. We aimed to systematically evaluate the effects of ITF intake on body weight, glucose homeostasis, and lipid profile on human subjects with different health status, including healthy, overweight and obese, prediabetes and diabetes, and hyperlipidemia. Weighted mean differences (WMDs) between ITF and control groups were calculated by a random‐effects model. A total of 33 randomized controlled human trials were included. Significant effect of ITF intake was only observed in the diabetics, but not in the other subject groups. Specifically, ITF intervention significantly decreased the WMD of blood glucose (−0.42 mmol/L; 95% CI: −0.71, −0.14; p = .004), total cholesterol (−0.46 mmol/L; 95% CI: −0.75, −0.17; p = .002), and triglycerides (TAG) (−0.21 mmol/L; 95% CI: −0.37, −0.05; p = .01) compared with the control. The stability of these favorable effects of ITF on diabetics was confirmed by sensitivity analysis. Also, ITF tends to lower LDL cholesterol (p = .084). But body weight and blood insulin were not affected by ITF intake. It should be noted that blood glucose, total cholesterol, and LDL cholesterol exhibited high unexplained heterogeneity. In conclusion, ITF intake lowers blood glucose, total cholesterol, and TAG in the people with diabetes, and they may benefit from addition of inulin into their diets, but the underlying mechanisms responsible for these effects are inconclusive.

ITF is a group of naturally occurring polysaccharides found in many plants (Roberfroid, 2005). They are classified according to their degree of polymerization (DP) into long-chain inulin (chain length of 11-60 DP with an average of 25 DP) and fructo-oligosaccharides (FOS) (chain length of 2-10 DP with an average of 4 DP) (Raninen et al., 2011;Roberfroid, 2007). Due to β (2,1) linkage, ITF is resistant to digestion in the small intestine and persisted into the colon where it undergoes fermentation. It is estimated that the Europeans and the Americans consume an average of 3-11 g and 1.3-3.5 g of ITF daily, respectively (van Loo et al., 1995). In 2018, inulin was approved by the Food and Drug Administration (FDA) in USA as an added dietary fiber so as to improve the nutritional value of the manufactured food products (Administration, 2018).
The data collected from animal models have indicated nutritional benefit of ITF in controlling obesity, diabetes, and hyperlipidemia, but only a few studies have been conducted to examine its effects on the human health (Beylot, 2005;Le Bourgot et al., 2018). Moreover, the data of human studies are not consistent, and thus, ITF benefit on human health is inconclusive. Therefore, the primary objective of this work was to comprehensively review the recent research performed on human subjects and carry out a meta-analysis evaluating the beneficial effect of ITF, if any, on humans. Parameters included in this analysis consist of changes to body weight, blood glucose, insulin, and lipid profile induced by ITF intake in human subjects. The secondary objective was to summarize the possible mechanisms underlying the beneficial effects of ITF intake observed in this study.

| Data sources and literature search
A comprehensive literature search for human intervention studies that evaluated the correlation between ITF intake and body weight, blood glucose or lipid profile, published between 1960 and 2020 was performed. NCBI, PubMed, Scopus, Ovid, EBSCO, Web of Science, ProQuest databases, Science, JSTOR, and MEDLINE were searched using the search terms 'inulin-type fructans', OR 'inulin' OR 'fructooligosaccharides' OR 'oligofructose' AND 'blood glucose' OR 'lipid' OR 'cholesterol' OR 'triglycerides' AND 'human', and the same terms were applied in each database during the search phase. In addition, a manual search of the reference lists of retrieved papers or review articles was conducted to identify all potentially relevant papers. No limit was placed on the language. Data were extracted by two reviewers. The present meta-analysis has been registered with PROSPERO (CRD42018117715).
The studies included in this review met the following criteria: (a) ITF intake exposure; (b) human subjects (≥18 years of age); (c) ran- an observational study, an animal study, a review or meta-analysis, a trial in the general population, a trial without relevant effect measures, or a non-ITF supplementation trial.

| Data extraction
The following data were extracted from each human study: lead author, year of publication, health status of subjects, number of subjects (female and male), BMI, age (year), number of subjects in the intervention and control groups, type of ITF consumed, dosage, duration, control, study design, and outcomes. The study duration in this review strictly referred to either the period of inulin intervention or the control diet rather than overall study period. For the missing data that were not explained in the corresponding articles, authors were contacted via email or phone call to seek permission for the data to be included. In cases where there is no answer, request refusal or data loss, the missing data were reported as "Not stated." The Heyland Methodologic Quality Score (MQS) was used to assess study quality (Heyland et al., 2001). Points were given on the basis of methodology (randomization, blinding, and analysis), sample (selection, compatibility, and follow-up), and intervention (protocol, co-intervention, and crossovers). A maximum of 13 points could be received. Study that received a score of 8 was considered to be of higher quality.

| Statistical analysis
Reviewer Manager (RevMan, version 5.3) was used to conduct the meta-analysis with changes in body weight, blood glucose, insulin, and lipid profile including total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides (TAG) as the outcomes. Heterogeneity across studies was quantified by using the I 2 statistic methodology, where each study design was considered, as a quantitative evaluation of inconsistency among the studies. To pool the results of studies with an acute impact on body weight, blood glucose, insulin, and lipid profile, a fixed-effects model was selected when heterogeneity was absent or low (I 2 < 20%), whereas when heterogeneity was greater, a random-effects model was utilized. In this work, weighted mean differences (WMDs) between treatment (ITF) and control groups were combined via a random-effects model to evaluate the size of treatment impacts. When the I 2 value is ≥20%, in which case the source of heterogeneity was explored by the removal of individual trials in the sensitivity analyses and through a priori subgroup analyses.
To examine whether a single study exerted undue influence on the overall results, sensitivity analyses were performed in which each individual study was excluded from the meta-analysis and the effect size was recalculated with the remaining studies. Meanwhile, a priori subgroup analyses were performed to further identify the possible sources of heterogeneity by comparing summary results obtained from subsets of studies grouped by characteristics (prediabetes or

| Literature search
As shown in Figure 1, 1,243 articles were identified and evaluated in the initial systematic search on the scientific databases.
Upon the removal of the duplicate articles (412) and articles that did not meet the eligibility criteria (810)

| Study characteristics
Extracted data from the human studies are summarized in Table 1.
All studies in this review were RCT studies, with 57% (19 trials) utilizing a parallel design, 40% (13 trials) utilizing a crossover design, and 3% (one trial) utilizing a Latin-square design. The subjects included in this analysis were stratified into healthy or one of the following metabolic symptomatic groups, including overweight and obese, prediabetes and diabetes, and hyperlipidemia. Of the 33 trials, the number of studies that had utilized healthy, overweight and obese, prediabetes and diabetes, and hyperlipidemia subjects as study objects was 8, 10, 10 and 5, respectively. Overall, ITF intake in RCT human studies ranged from 3 to 30 g/day of ITF as part of the ingredient in the diet (11 g, median levels of individual series). ITF was provided as beverages, ice cream, or natural solid foods such as cookies, pasta, and bread rolls. Treatment duration ranged from 2 to 18 weeks with the median length of 6 weeks. Blood samples were obtained at baseline and after diet intervention. Other parameters including body weight, blood glucose, insulin, or lipid profile were collected.

| Effect on body weight
To assess the intake of ITF in assisting with weight loss, a total of 17 studies were included in this analysis (Figure 2). The result showed that ITF intake had no effect on the body weight in the overall analysis (WMD, −1.51 kg; 95% CI: −3.89, 0.87; p = .21) and any subgroups (p > .05). There was a significant intertrial heterogeneity in the overweight and obese subgroup (I 2 = 64%, p = .004), as well as borderline significance in intertrial heterogeneity in the overall pooled analysis (I 2 = 39%, p = .05).

| Effect on blood glucose
Results for blood glucose were reported in 24 eligible studies ( Figure 3).
Overall, ITF led to lower glucose concentration when compared to a

| Effects on insulin
Overall, significant reduction in insulin concentration was observed

| Effects on total cholesterol
Overall, a significant reduction in total cholesterol (WMD, −0.15 mmol/L; 95% CI: −0.29, −0.02; p = .03) was observed in 26 trials ( Figure 5). The stratified analysis found that only prediabetes and diabetes subjects showed a significant reduction in total cholesterol concentration after ITF-supplemented diet (WMD, −0.46 mmol/L; 95% CI: −0.75, −0.17; p = .002). This is based on a total of 8 trials, which accounted for almost one quarter of weight (24.3%) of the overall analysis. The overall test for heterogeneity resulted in I 2 = 36% (p = .04). Heterogeneity for total cholesterol was markedly reduced when studies on prediabetes and diabetes subjects were removed from the analysis (I 2 = 11.8%, p = .31). The data indicated that this group was largely responsible for the observed heterogeneity.

| Effect on LDL cholesterol
In the 23 studies that evaluated LDL cholesterol, there was evi-
The stratified analysis found that only studies performed on patients with prediabetes and diabetes, which accounted for almost one third of the weight (34.8%), showed reduction in HDL cholesterol concentrations (WMD, 0.07 mmol/L; 95% CI: 0.01, 0.12; p = .02; Figure 7).
No evidence of heterogeneity in the overall analysis and all subgroups was detected with I 2 = 0% (p > .05).

| Effect on triglyceride
In this investigation, the borderline significance (WMD, −0.07 mmol/L; 95% CI: −0.14, 0.00; p = .05) effect of ITF intake on TAG was observed between ITF and control group as showed in a total of 26 studies, with little heterogeneity (I 2 = 7%, p = .37;

| Publication bias
Next, publication bias of these trials was examined by funnel plot analysis, Begg's test, and Egger's test ( Figure S1). With exception to body weight and blood glucose, visual inspection of the funnel plots found that individual studies of WMD estimates were reasonably symmetrical. The absence of publication bias (p > .05) was supported by both Begg's test and Egger's test. For body weight, visual inspection of funnel plot indicated potential asymmetry, this was, however, not confirmed by Begg's test (p = .149) and Egger's test (p = .578).
For trials that report blood glucose, statistical evidence of publication bias (Begg's test, p = .018; Egger's test, p = .006) was observed.

| Side effects
Detailed information gathered from the 33 RCTs in this review showed that ITF was mostly well tolerated by subjects of different health status. However, minor side effects such as abdominal flatulence, bloating, and nausea were reported (Daud et al., 2014;de Luis et al., 2013;Guess et al., 2014). None of these effects were recognized as adverse to the human health, and the symptoms eventually subsided with diet adaptation over time (Parnell & Reimer, 2009).
Nonetheless, it has been reported that adding 20 g FOS/70 kg body weight/day led to significant gastrointestinal side effects, which is not seen if 10 g/day was offered (Genta et al., 2009). Scientifically, from the view of gut microbiota modulation, slight flatulence and bloating are probably positive cues that the prebiotics is effective.
Fermentation of prebiotics by the gut microbiota in the colon gave rise to the production of gas and acid, resulting in the symptoms.

| D ISCUSS I ON
ITF supplementation in diet has been suggested to alleviate several features of metabolic syndrome including obesity, diabetes, and F I G U R E 2 Forest plot of randomized controlled trials investigating the effect of ITF intake on body weight (kg) compared with control groups for human studies. Results from individual trials were pooled with random-effects model and are expressed as weighted (squares) mean differences with 95% CIs (horizontal lines). Pooled effect estimates (diamonds) are presented for each subgroup as well as the overall analysis hyperlipidemia; however, results from human trials remained inconsistent. This current study demonstrates that the favorable outcome of ITF intake was observed only in prediabetes and diabetes subjects, and favorable outcome was defined as significant blood glucose, total cholesterol, and TAG concentration reduction after study duration. Current data indicate that ITF only benefits individual of certain health status. Also, the removal of subjects with prediabetes did not change the significance of effects on blood glucose and lipid profiles; instead, increased significance was observed (data not shown).
Our current results are, in part, in line with a recent metaanalysis where dietary prebiotic intake lowered the plasma TC, LDL-C, and TAG concentrations of the diabetes trials included in their analysis (Beserra et al., 2015). The magnitude of the treatment response varies depending on the pathological state and the gut microbiota composition (Beserra et al., 2015;Larsen et al., 2010;Wu et al., 2010). Notably, a previous meta-analysis by Liu et al. (2017) showed that ITF intake is beneficial by sustaining glucose homeostasis and reducing LDL-C level; yet, there are several differences between Liu's meta-analysis and ours. Firstly, in Liu's F I G U R E 3 Forest plot of randomized controlled trials investigating the effect of ITF intake on blood glucose (mmol/L) compared with control groups for human studies. Results from individual trials were pooled with random-effects model and are expressed as weighted (squares) mean differences with 95% CIs (horizontal lines). Pooled effect estimates (diamonds) are presented for each subgroup as well as the overall analysis meta-analysis, both studies by Dehghan et al. (2013)

| Effect on body weight
There are increasing epidemiological studies, suggesting that dietary-fiber-rich diets are linked to lower body weight or BMI (Du et al., 2010;Grube et al., 2013;Thompson et al., 2017). Evidence from several animal studies has shown that ITF supplementation lowers energy intake through diet, promotes weight loss, and improves body composition like reduction in fat mass (Arora et al., 2012;Cani et al., 2004;Cani, Neyrinck, et al., 2005;Delmée et al., 2006;Dewulf et al., 2011). Above studies have speculated that fermentation of the ITF in the gut resulted in the higher concentration of peptide YY (PYY) and glucagon-like peptide 1 (GLP-1) via the short-chain fatty acids (SCFA). Consistently, independent of other lifestyle changes, a 12-week treatment with 21 g/day FOS in subjects who are overweight and obese has been shown to decrease energy intake through diet and body weight. This is likely due to suppressed ghrelin and F I G U R E 4 Forest plot of randomized controlled trials investigating the effect of ITF intake on plasma or serum insulin (µlU/ml) compared with control groups for human studies. Results from individual trials were pooled with random-effects model and are expressed as weighted (squares) mean differences with 95% CIs (horizontal lines). Pooled effect estimates (diamonds) are presented for each subgroup as well as the overall analysis enhanced PYY, but not GLP-1, which remained unchanged (Parnell & Reimer, 2009). However, based on the results from this metaanalysis, it appeared that the beneficial effects of ITF intake were not associated with weight loss. This is probably due to the fact that most studies did not include restricted energy intake through diet or intention to lose weight as one of the factors in their data analysis.

| Effect on glucose
The current meta-analysis showed that ITF intake significantly decreased blood glucose concentration in prediabetes and diabetes subjects. Evidently, this beneficial effect was also supported by several diabetes animal model studies (Bharti et al., 2013;Byung-Sung, 2011;Cani, Daubioul, et al., 2005;Cani et al., 2006;Gobinath et al., 2010;Ning et al., 2017;Zhang et al., 2018). Due to the lack of viscosity property, ITF is not associated with relatively low absorption in the gut.
Despite this, there are several mechanisms that may be responsible for the effect of ITF on blood glucose. It is generally recognized that ITF, via the fermentation by the intestinal microbiota, produces high level of SCFA end products. Of which, 90%-95% comprised of acetate, butyrate, and propionate. Butyrate directly increases the PYY/ proglucagon (the gene that encodes GLP-1) gene expression both in vitro and in vivo in rat study (Zhou et al., 2006). Accordingly, it had F I G U R E 5 Forest plot of randomized controlled trials investigating the effect of ITF intake on total cholesterol (mmol/L) compared with control groups for human studies. Results from individual trials were pooled with random-effects model and are expressed as weighted (squares) mean differences with 95% CIs (horizontal lines). Pooled effect estimates (diamonds) are presented for each subgroup as well as the overall analysis been reported that FOS significantly promotes the expression and secretion of colonic GLP-1 amide (7-36), and the antidiabetic impact of FOS is dependent on the action of GLP-1 in the streptozotocininduced diabetic rats (Cani, Daubioul, et al., 2005;Cani et al., 2006).
Consistently, Zhao et al. (2018) also found that increased fecal bu- Here, propionate depletes hepatic citrate, which is the main inhibitor of one of the most important regulatory enzymes of glycolysisphosphofructokinase (Roberfroid & Delzenne, 1998).

| Effect on cholesterol
In the medical field, treatments aimed to decrease cholesterol concentration are effective ways of combating the risk of cardiovascular diseases (Lloyd-Jones et al., 2010;Yusuf et al., 2004). The cholesterol-lowering effects of ITF observed may be accounted for by several mechanisms. Firstly, ITF mediates in vivo bile acid level. In inulin-fed rats, lower serum cholesterol concentration was accompanied by an increase in fecal bile acid and neutral steroid excretion F I G U R E 6 Forest plot of randomized controlled trials investigating the effect of ITF intake on LDL cholesterol (mmol/L) compared with control groups for human studies. Results from individual trials were pooled with random-effects model and are expressed as weighted (squares) mean differences with 95% CIs (horizontal lines). Pooled effect estimates (diamonds) are presented for each subgroup as well as the overall analysis (Han et al., 2013;Levrat et al., 1994;Parnell & Reimer, 2010). The loss of bile acids in the feces enhanced cholesterol uptake by the liver from the circulation to replenish the bile acid supply. It has been reported that fecal bile acid excretion is inversely correlated with liver cholesterol concentrations (r 2 = .20) (Vanhoof & De Schrijver, 1995).
Consistently, Yang and colleagues further suggested that the suppressive effect of ITF on cholesterol was mediated by the inhibition of cholesterol de novo synthesis as evidenced by the reduced Srebf2 and Hmgcr gene expression. Due to its low viscosity, ITF does not bind to the bile acids in the intestinal lumen (Schneeman, 1999). But intestinal pH is lowered as the result of the organic acids produced during ITF fermentation in the intestinal mucosa. Consequently, the bile acids become less soluble and excreted with the feces, thereby reducing their intestinal absorption (Pereira & Costa, 2002).
Secondly, as discussed previously, both acetate and propionate enter the liver through the portal vein. In rat administrated with FOS, both acetate and propionate concentrations risen more than twofold in the portal serum of the rats (Roberfroid & Delzenne, 1998

| Effect on triglyceride
In rats, mice, and hamsters fed with ITF-supplemented diet, hepatic lipid metabolism was regulated with reduced TAG accumulation in the hepatic and/or reduced serum lipids. Several mechanisms were proposed. Firstly, studies in the animal models clearly showed that lowered hepatic and/or serum TAG concentration were mainly due to the inhibition of TAG-rich VLDL particle secretion (Delzenne et al., 2007;. It has been proposed that the hypotriglyceridemic action is probably attributed to the downregulation of de novo fatty acid synthesis in the liver, as evidenced by 50% decrease in the key hepatic lipogenesis enzyme activities, such as acetyl-CoA carboxylase, fatty acid synthase, and malic enzyme. This is also reflected by a substantial reduction in fatty acid synthase mRNA, supporting the hypothesis that FOS treatment can modify lipogenic enzyme gene expression Delzenne & Kok, 2001).
Secondly, there were evidences suggesting that ITF may prevent the F I G U R E 8 Forest plot of randomized controlled trials investigating the effect of ITF intake on triglycerides (mmol/L) compared with control groups for human studies. Results from individual trials were pooled with random-effects model and are expressed as weighted (squares) mean differences with 95% CIs (horizontal lines). Pooled effect estimates (diamonds) are presented for each subgroup as well as the overall analysis esterification step, in which hepatocytes isolated from FOS-fed rats showed decreased capacity to esterify 14 C-palmitate and 14 C-acetate into TAG by 40% and 54%, respectively (Fiordaliso et al., 1995;Kok, Roberfroid, Robert, et al., 1996). However, the high levels of fat present in most human diets mean that the rate of de novo fatty acid synthesis in the liver is extremely low, as the exogenous dietary fatty acids provided all the required substrates for triglyceride VLDL synthesis. Thus, the explanation above cannot be extrapolated to humans (Aarsland et al., 1996). Thirdly, there is a possibility that ITF reduces serum TAG through an extrahepatic mechanism, namely through triglyceride-rich lipoprotein catabolism enhancement (Daubioul et al., 2000;Delzenne et al., 2002). In one study, ob/ob rat fed with FOS displayed lower plasma TAG and muscle lipids. The phenomenon can be ascribed to 70% increase in the lipoprotein lipase mRNA expression in the muscle tissue (Everard et al., 2011).

| Limitations
No study is without limitations. Firstly, the limited number of studies conducted to date, combined with small sample sizes and short intervention periods, insufficiently powered to support the effect, thus limiting the generalizability of observed effects on blood glucose and lipid profiles to larger populations with ITF intake. Secondly, there was considerable variability in study design between studies within subgroup of prediabetes and diabetes included in the metaanalyses, with different study durations and dose of ITF consumed.
These effects should be taken into account in meta-regression analysis, but testing the number of studies available for such analyses is still inadequate. Finally, gap remained to fully elucidate the mechanisms underlying the protective effects on blood glucose and lipid profiles in subjects with prediabetes and diabetes.

| CON CLUS ION
In conclusion, the present meta-analysis proposed that increased intake of ITF significantly reduces blood glucose, total cholesterol, and TAG in subjects with prediabetes and diabetes, without affecting the other subject groups. Because of the results of our study and the findings from previous studies, patients with prediabetes and diabetes may benefit from inulin supplementation in the reduction in blood glucose and the amelioration of the underlying health conditions. Several mechanisms were proposed to explain each health benefit that occurred in patients who are prediabetes and diabetes, but they remain inconclusive. Further research, especially large, well-powered, long-term human intervention studies, is required to further understand and promote the role that ITF plays in human health management.

ACK N OWLED G M ENTS
The authors' responsibilities were as follows: LX, LKL, and LP designed research; LX and LKL conducted research; LKL analyzed data and wrote the paper; LX had primary responsibility for final content. None of the authors declared a conflict of interest. All authors have read and approved the final manuscript. This work was supported by grants from the National Key R&D program of China (2019YFA0801701 to LX), National Science Foundation of China (91854104 to LX, 31690103 to PL), and National Key R&D program of China (2018YFA0506901 and 2016YFA0502002 to PL).

CO N FLI C T O F I NTE R E S T
There are no conflicts of interest to declare.

AUTH O R CO NTR I B UTI O N S
Liangkui Li: Formal analysis (lead); Methodology (equal); Resources

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author.