点击显示 收起
1 From the Departments of Gastroenterology (YB) and Bacteriology (LR) and the Therapeutic Research Unit (GS), Hôpital Lariboisière, Paris; the Clinical Research Unit, Fernand Widal Hospital, Paris (EV); the Faculty of Pharmacy, Lille, France (CN); the Department of Gastroenterology, Lyon South Central Hospital, Pierre Bénite, Lyon, France (BF); Cerestar R&D Center, Vilvoorde, Belgium (FB); Nutrition & Toxicology Research Institute, Maastricht University, Maastricht, Netherlands (FB); and Nutri-Health, Rueil-Malmaison, France (FRB)
2 Supported by a grant from the Health & Nutrition Group, Vilvoorde, Belgium. 3 Address reprint requests to Y Bouhnik, Hôpital Lariboisière, 2, rue Ambroise Paré, 75475 Paris Cedex 10, France. E-mail: yoram.bouhnik{at}lrb.ap-hop-paris.fr.
ABSTRACT
Background: Nondigestible carbohydrates (NDCHs) are fermented in the colon, where they can selectively promote the growth of bifidobacteria.
Objective: Our aim was to determine the bifidogenic potential of different NDCHs used in human diets.
Design: Two hundred healthy volunteers participated in this double-blind study. During phase 1 (screening), 64 volunteers were randomly assigned to 8 groups of 8 subjects each; for 7 d, they ingested 10 g/d of 1 of the 7 NDCHs tested or of the placebo. During phase 2 (dose-response study), 136 volunteers were randomly assigned to 4 groups of 32 subjects who received 2.5, 5.0, 7.5, or 10 g/d, respectively (8 subjects/dose), of one of the NDCHs that were proven to be bifidogenic during phase 1 and a fifth group of 8 subjects (control subjects) who received the placebo. Stools were recovered before and after NDCH consumption.
Results: In phase 1, 4 NDCHs were found to be bifidogenic: short-chain fructooligosaccharides (P = 0.008), soybean oligosaccharides (P = 0.006), galactooligosaccharides (P < 0.0001), and type III resistant starch (P = 0.02); lactulose, long-chain inulin, and isomaltooligosaccharides were not bifidogenic. In phase 2, the effects of 7-d treatment on bifidobacteria concentrations were found to differ significantly among the 4 NDCHs (P = 0.009 for time x treatment interaction). However, no significant differences were found among doses, and there was no significant dose x time interaction. A low baseline bifidobacteria count was significantly associated with the bifidogenic response to treatment (P < 0.001).
Conclusion: This study showed the different bifidogenic properties among the substrates and underlined the importance of taking into account the baseline bifidobacteria counts when evaluating the effect of the treatment.
Key Words: Bifidobacterium • human fecal microflora • gut flora • nondigestible carbohydrate • oligosaccharides • prebiotics • randomized controlled trial
INTRODUCTION
The human large gut contains a wide variety of bacterial genera, species, and strains that are thought to be either beneficial (eg, Bifidobacterium and Lactobacillus) or detrimental (eg, Clostridium) to the host's health (1, 2). Although this generalization probably is too simplistic a view of gut microbiology, it is a feasible concept from which to develop food components that function to modulate the composition of the colonic microbiota (3). It is in this context that a prebiotic was defined as "a nondigestible food ingredient that beneficially affects the host by selectively stimulating the growth and/or the activity of one or a limited number of bacterial species in the colon" (4). Whereas a probiotic introduces exogenous bacteria into the colonic microbiota, a prebiotic aims at stimulating the growth of one or a limited number of the potentially health-promoting indigenous microorganisms, thus modulating the composition of the natural ecosystem. Of the currently available food ingredients, carbohydrates that resist digestion in the upper gastrointestinal tract (nondigestible carbohydrates, NDCHs) but that are hydrolyzed and fermented in the large bowel are the only known components for which convincing evidence in favor of a prebiotic effect has been reported.
Over the last decade, several NDCHs have been investigated for their effects on the colonic microflora. It has emerged that some NDCHs, mainly the oligosaccharides, have the potential to increase bifidobacteria concentrations in the colon (5). The ability to act as a prebiotic provides a marketing edge for these products, which has encouraged research into the ability of carbohydrates to induce beneficial changes in the composition and metabolism of the colonic microflora (6). Comparisons of in vitro fermentation properties of commercial prebiotic oligosaccharides showed that all prebiotics increase the number of bifidobacteria, and most reduce the number of clostridia (7). However, studies in humans have been limited (8). The aims of this study were to ascertain the bifidogenic properties of different NDCHs used in human diets, to search for dose-response relations for the bifidogenic NDCHs, and to evaluate possible gastrointestinal side effects associated with the consumption of these NDCHs in healthy humans.
SUBJECTS AND METHODS
Subjects
Two hundred healthy volunteers (81 men and 119 women) aged 18–54 y (mean age: 30 y) participated in the study. Exclusion criteria were a history of gastrointestinal disease other than appendicitis, antibiotics or laxatives taken during the 2 mo before the study, and other medication taken during the investigation period. Subjects gave written informed consent, and the protocol was performed in accordance with the guidelines of the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use and the principles laid down in the current revision of the Declaration of Helsinki, which was approved by the Consultative Committee for Protection of People during Biomedical Research (Paris-St-Louis, France).
Testing of nondigestible carbohydrates
The NDCHs tested were short-chain fructooligosaccharides [(sc-FOS) Actilight; Beghin Meiji, Paris], soybean oligosaccharides (Calpis Food industry, Tokyo), galactooligosaccharides (Cup-oligo P: Niossin Sugar MFG, Tokyo), isomaltooligosaccharides [(IMO) Cerestar, Vilvoorde, Belgium], debranched retrograded tapioca maltodextrin [(a type 3 resistant starch; RS) Actistar, Cerestar], long-chain inulin (Orafti; Raftiline HP, Tienen, Belgium) and lactulose (MLC-A; Meiji Seika Kaisha, Tokyo). The placebo was 50% sucrose and 50% fully digestible waxy maize-derived maltodextrins (DE6.5; Cerestar). Each product contained >95% of the stated material and was not contaminated by any other monosaccharides or disaccharides.
Experimental design
The study had 2 parts—phase 1 consisted of a prescreening study, and phase 2 consisted of a dose-response study. Phase 1 evaluated the effect on flora of all selected NDCHs when consumed at a dose of 10 g/d. A placebo group was composed to show the relative stability of bifidobacteria concentrations over time and to be used as a comparison for all the NDCHs tested. An NDCH was considered to be bifidogenic when its consumption resulted in a significantly greater fecal bifidobacteria count than did consumption of placebo.
All subjects consumed their usual daily diet from the preinclusion day (day 0) to the end of the study (day 15). They were instructed to exclude food products containing any of the NDCHs under study. Fermented dairy products containing viable bifidobacteria were not allowed, because it was already shown that their consumption could lead to a rise in fecal bifidobacteria counts within a few days (9).
During phase 1, 64 volunteers were randomly assigned to 8 groups of 8 subjects. From day 8 to day 14, they ingested 5 g of one of the 7 NDCHs or placebo after both lunch and dinner. Phase 2 aimed to evaluate possible dose-response effects on the flora of the NDCHs that had been found during phase 1 to be bifidogenic (at a dose of 10 g/d). For this purpose, 136 volunteers were randomly assigned to 5 groups: 4 groups of 32 subjects and a placebo group of 8 subjects, which was used for evaluating digestive symptoms. Each group of 32 subjects was further divided randomly so that subgroups of 8 subjects ingested a daily dose of 2.5, 5.0, 7.5, or 10 g/d from day 8 to day 14 divided between 2 oral doses, one dose consumed after lunch and one after dinner.
Digestive symptoms
Gastrointestinal side effects were evaluated during phases 1 and 2 by using a daily chart on which the symptoms (ie, excess flatus, bloating, borborygmi, and abdominal pain) were graded from 0 (no symptom) to 3 (severe symptoms), as described previously (10, 11). Frequency and consistency of stools were also noted, and diarrhea was defined as 1 watery stool or >3 stools/d.
Stool collection
Stools were recovered twice, on day 8 (corresponding to the first day before the start of the NDCH consumption) and on day 15 (ie, after 7 d of NDCH consumption, which lasted from day 8 to day 14). Samples were collected in plastic containers that had been rendered anaerobic (Anaerocult A; Merck, Darmstadt, Germany); the containers were immediately transferred to the laboratory, and the samples were analyzed immediately. The time from stool emission to bacteriologic analysis was <1 h.
Bacteria counts and pH
Fecal samples (1 g) were introduced into the first tube of the dilution series (which had been weighed beforehand) and thoroughly mixed and then further decimal dilutions were made up to –9 in a reduced diluent (cysteinated [1/4]-strength Ringer solution diluent). We spread 0.1 mL of each dilution onto plates with different selective media to outnumber several bacterial genera: total anaerobic counts (Wilkins-Chalgren agar), Bifidobacterium (Beerens's medium), Bacteroides (Bacteroides Bile Esculin agar), Lactobacillus (Lactobacillus agar according to De Man, Rogosa, and Sharpe; MRS agar), and enterobacteria (McConkey agar). The tests were duplicated for the first 2 media. Plates of the first 3 media were incubated anaerobically for 5–7 d, those for the MRS agar were incubated for 48 h under atmosphere enriched in CO2, and those for McConkey agar were incubated aerobically for 48 h. Colony counts were obtained and expressed as a log of the colony-forming units (CFUs)/g fresh feces. The fresh stool pH was immediately measured by using a pH meter (Bioblock, Illkirch, France).
Data analysis
Bacteriologic analysis
In phase 1 of the study, we used a two-factor analysis of variance for one within-factor variable (ie, time) and one between-factor variable (ie, treatments), which allowed us to test the hypotheses of overall differences between treatments and times and to ascertain whether the changes in bifidobacteria counts observed during the 7-d treatment differed significantly between treatments. Individual comparisons among treatments were made only when the overall P value was < 0.05.
Treatments significantly different from placebo were selected for phase 2 of the study. Because phase 1 was a screening phase, no adjustment of significance level for multiplicity was used for the selection of the treatments, but P values adjusted by Bonferroni's method are presented in Table 1.
View this table:
TABLE 1. Fecal bifidobacteria and total anaerobe counts in 7 groups of 8 healthy volunteers during a 7-d consumption of various nondigestible carbohydrates (NDCHs) at a dose of 10 g/d and in 8 subjects consuming placebo1
In phase 1 of the study, we used a three-factor analysis of variance for one within-factor variable (ie, time) and 2 between-factor variables (ie, treatment and dose effect), which allowed us to test the hypotheses of overall differences among treatments, times, and dose effects. Once more, individual comparisons were made only when overall P value was < 0.05. The Bonferroni adjustment for multiplicity was used for individual comparisons. Logarithmic transformation was used before analysis of variance for normalization and variance stabilization purposes.
Digestive tolerance
The occurrence and the intensity of digestive symptoms experienced in conjunction with the consumption of NDCHs were compared with those after the consumption of placebo by using the Bonferroni test. Digestive symptom intensity was assessed as a 7-d score before and after treatment. Intragroup analysis was tested by using Wilcoxon's signed-rank test.
RESULTS
Fecal bacterial counts and pH
Phase 1
A significant difference among the 8 treatments tested was found with respect to the change in bifidobacteria concentrations during the 7-d consumption of NDCHs (P < 0.0003, = 0.003, and < 0.001 for treatment, time, and time x treatment interaction, respectively). Four NDCHs were found to be statistically different from placebo—ie, they were bifidogenic at 10 g/d: sc-FOS, soybean oligosaccharides, galactooligosaccharides, and type 3 RS; lactulose, long-chain inulin, and IMO were not bifidogenic (Table 1). During the study, neither fecal pH nor counts of total anaerobes, Lactobacillus, Bacteroides, or enterobacteria changed in any of the groups (data not shown).
Phase 2
No significant differences in baseline values were found among the different groups (P = 0.34). The effects of the 7-d treatment were found to differ significantly among the 4 NDCHs (P = 0.009 for time x treatment interaction). However, no significant differences were found among doses tested (dose and dose x time interaction were NS). Interaction between the 3 factors (time x dose x treatment) was equal to 0.06, which suggests that the relation between doses and changes in bifidobacteria counts during the 7-d treatment can differ significantly among the 4 NDCHs (Table 2). When, for exploratory purposes, we considered the relations between bifidobacteria and doses for the 4 NDCHs tested, we found a linear dose-response relation from 2.5 to 10 g/d for sc-FOS (P < 0.001) but not for the other NDCHs.
View this table:
TABLE 2. Fecal bifidobacteria and total anaerobe counts in 128 healthy volunteers assigned to a 7-d consumption of 4 bifidogenic nondigestible carbohydrates (NDCHs) at a dose of 2.5 to 10 g/d1
Note also that there was a significant relation between treatment effect and initial bifidobacteria count on day 8 (P < 0.001). In addition, when the 4 NDCHs were compared on day 15 (after 7 d of treatment), a significant difference was found among them (P = 0.02). Two-by-two NDCH comparisons reached significance (adjusted for multiplicity) only between type 3 RS and galactooligosaccharides (mean values were 9.8 ± 0.21 and 10.42 ± 0.20, respectively; P = 0.05); there was no difference among sc-FOS, soybean oligosaccharides, and galactooligosaccharides.
Digestive tolerance
Phase 1
Significant increases in excess flatus, bloating, borborygmi, and abdominal pain were observed during the 7-d consumption of NDCHs, but no significant difference among the 8 treatments was found with respect to changes in digestive symptoms during the consumption of NDCHs (P < 0.02 for time and P > 0.15 for time x treatment interaction for these 4 variables; Table 3). No significant effect was found for the number of stools. Diarrhea was not reported in any of the groups.
View this table:
TABLE 3. Digestive symptom intensity in 7 groups of 8 healthy volunteers during a 7-d consumption of different nondigestible carbohydrates (NDCHs) at a dose of 10 g/d compared with 8 subjects given placebo1
Phase 2
Significant increases in excess flatus, bloating, and abdominal pain were observed during the 7-d consumption of NDCHs, but no significant differences were found among the 4 treatments tested with respect to changes in digestive symptoms during the consumption of NDCHs (P < 0.01 for time and P > 0.45 for time x treatment interaction for these 3 variables; Table 4). No significant effect was found for borborygmi or the number of stools. Diarrhea was not reported in any of the groups.
View this table:
TABLE 4. Digestive symptom intensity in 4 groups (n = 32 per group) of healthy volunteers during a 7-d consumption of different nondigestible carbohydrates (NDCHs) at various doses from 2.5 to 10 g/d compared with 8 subjects given placebo1
DISCUSSION
This double-blind, randomized, placebo-controlled, parallel-group study showed that sc-FOS, soybean oligosaccharides, galactooligosaccharides, and type 3 RS were bifidogenic at a dose of l0 g/d when consumed for 8 d by humans, whereas long-chain inulin, lactulose, and IMO were not. Among the 4 NDCHs tested, the bifidogenic effect was stronger for soybean oligosaccharides and galactooligosaccharides than for sc-FOS and type III RS. Our study also confirmed that results obtained in vitro by using selected strains or mixed human inoculate [ie, almost all NDCHs increased the numbers of bifidobacteria, and most decreased clostridia (7)] cannot readily be generalized as being valid also for the in vivo situation. Thus, the food matrix and the complex ecosystem in vivo may affect the fermentation pattern and influence the outcome.
The importance of a placebo and the influence of basal bifidobacteria counts on the response to NDCH consumption
In most studies evaluating the bifidogenicity of certain substrates, linear study designs are used and measurements are made before and after treatment (12–15) without the inclusion of subjects receiving a placebo (ie, a control group). Our results indicate the importance of a placebo treatment to exclude the possible effects of time and unknown environmental factors. The effect of a control group was also underlined in the study of Alles et al (16), in which the contrast between the control group and the study groups determined their conclusion as to the bifidogenicity of the substrate tested. One of the possible confounding factors in interpreting an increase in bifidobacteria concentration is the initial bifidobacteria count. We confirmed that a low baseline bifidobacteria count was significantly associated with an increased count after treatment.
The duration of nondigestible carbohydrate administration
We chose a relatively short-term NDCH administration (7 d) because previous studies showed that this time was long enough to lead to significant modifications in bifidobacteria counts by using different substrates (11, 14, 17, 18). Other studies evaluated the bifidogenic effect of substrates after longer administration (13, 16, 19, 20), and we cannot exclude the possibility that substrates found in the present study to be nonbifidogenic would have been found to be bifidogenic if longer administration had been used.
Bifidogenic nondigestible carbohydrates
NDCHs that were found to be bifidogenic in phase 1 were sc-FOS, soybean oligosaccharides, galactooligosaccharides, and RS. Sc-FOS has been studied extensively, and its bifidogenic effect was shown in well-controlled human trials (10, 11, 13, 14). In the current study, we found a linear dose-response relation at doses from 2.5 to 10 g/d, which suggests a dose-effect relation (11). We observed that soybean oligosaccharides also increased bifidobacteria counts at all doses from 2.5 to 10 g/d, which was previously shown in humans only at a dose of 10 g/d (17). Galactooligosaccharides increased bifidobacteria counts in all doses tested, from 2.5 to 10 g/d. A bifidogenic effect of galactooligosaccharides was previously observed in 2 linear design studies at a daily intake of 10 g/d for 1 wk (15, 18), but not in a placebo-controlled study that evaluated doses of 7.5 and 15 g/d (16). In the current study, the first one performed in humans, we found that type 3 RS was bifidogenic at a dose of 10 g/d in phase 1. However, although bifidobacteria counts increased at doses of 5 and 7.5 g/d, they decreased at doses of 2.5 and 10 g/d, and thus no firm conclusion can be drawn about the bifidogenic properties of this substrate. Thus far, the few data available on the bifidogenic effects of type 3 RS are from studies in rats and in vitro studies using mixed human inoculate (21, 22).
Except for sc-FOS, we did not find any dose-response relation. This relation was previously found by using sc-FOS, but the range of doses was wider, from 0 to 20 g/d (11). In the present study, the range of different doses tested was narrow, from 2.5 to 10 g/d, and when we take into account the low accuracy of bacteriologic counting methods in stool, we cannot exclude that that low accuracy was due to a beta risk.
Nonbifidogenic carbohydrates
Unlike previous studies (13, 23), the current study found that long-chain inulin was not bifidogenic at a dose of 10 g/d. This negative result could be explained by the high dose and long duration of inulin administration in previous studies. In fact, a bifidogenic effect of inulin was previously reported in humans at doses ranging from 15 to 40 g/d (13, 23). A direct comparison of inulin to fructooligosaccharides [GFn = 2 <n < 6 (where G is glucose, F is fructose, and n is between 2 and 6)] in a human feeding trial found that inulin had a prebiotic effect similar to that of fructooligosaccharides (13). However, that study should be interpreted with caution because only 4 subjects were studied. In another study, inulin was administered to 10 elderly subjects, initially at a dose of 20 g/d and thereafter increasing to 40 g/d (23).
In the present study, lactulose was not bifidogenic at a dose of 10 g/d, although there was a trend (P = 0.08). Lactulose was previously reported to be bifidogenic in humans at doses of 20 g/d (24), 10 g/d (19) and even as low as 3 g/d (25). Our current negative result may be due to the short administration (8 d), whereas other studies lasted from 2 to 6 wk (19, 25). IMO was not bifidogenic at the dose of 10 g/d. This substrate was previously reported as bifidogenic in humans, but only in one open, noncontrolled study (26).
Effects of NDCHs on other intestinal bacteria
There was no effect of NDCHs on total anaerobes. Using sc-FOS, we did not find any significant reduction in any other genus, whereas Gibson et al (13) found a significant reduction in Bacteroides by using fructooligosaccharides. The reason for these discrepancies is unclear. As with sc-FOS, we did not observe any effect of soybean oligosaccharides, galactooligosaccharides, lactulose, type 3 RS, or IMO on the other intestinal bacteria. No published data are available for comparison.
Fecal pH
Fecal pH remained unchanged throughout the studies. A decrease in colonic pH might reduce the risk of developing colonic cancer, because an inverse correlation between stool pH and colon cancer risk was observed (27, 28). A slight acidification of fecal contents during sc-FOS consumption has been observed in animals (29) and in humans (30). In our previous study (10), fecal pH did not change during the consumption of sc-FOS or inulin. However, because the fecal pH is the net sum of the degree of short-chain fatty acid absorption and bicarbonate secretion during passage through the colon, fecal pH does not reflect the pH in the colon under physiologic conditions (31, 32).
Digestive tolerance
Symptoms relating to gas production in the gut are widely reported in human prebiotic feeding studies, but they remain very mild at recommended intakes (13, 33). Whereas there were significant symptoms of digestive intolerance in the placebo group, we did not find such symptoms in any NDCH group during phase 1 and found only minor bloating in sc-FOS, soybean oligosaccharides, and type 3 RS during phase 2. However, the decrease in symptoms in the placebo group during the treatment period may explain the latter finding. These results underline the necessity of including a placebo group as a control for evaluating subjective digestive symptoms.
In summary, this study is the first that has systematically evaluated the effects of several NDCHs that are frequently used in human diets. Its design, including a large population of healthy volunteers and optimal conditions for bacteriologic analysis, showed the different bifidogenic properties among the substrates and the effect of baseline bifidobacteria counts on the degree of response.
ACKNOWLEDGMENTS
The authors' contributions to the study and this report were as follows: study design (YB, FB, and FRB), data collection (YB, LR, and GS), data analysis (YB, GS, EV, and FB), review of data (EV), test substance preparation (FB), and writing of the manuscript (YB, CN, BF, FB, and FRB). FB is employed at the Nutrition & Toxicology Research Institute at Maastricht University and at Cerestar R&D Center, and FRB is employed at Nutri-Health. No other authors had any conflicts of interest.
REFERENCES