In human nutrition, fiber-rich diets are sometimes considered as a means to reduce body weight or to prevent obesity (Blundell and Burley 1987, Rigaud et al. 1987). Food manufacturers use dietary fiber [non-starch polysaccharides (NSP) plus lignin] and other poorly digested carbohydrates (e.g., Polydextrose^R and inulin) as constituents of low energy foods where they replace full-energy nutrients such as sucrose or fat. Although dietary fibers are not hydrolyzed in the small intestine by mammalian digestive enzymes, parts of their energy can become available by microbial breakdown of NSP in the large intestine. The main products of fermentation, the short-chain fatty acids (SCFA), are nearly completely absorbed (McNeill et al. 1978) and can be used as fuels for the tissues of the host (Rémésy et al. 1992).

An increased intake of dietary fiber is associated with a concomitant loss of energy to feces. Besides the higher energy loss caused by unfermented dietary fiber, an increased intake of dietary fiber is associated with additional losses of energy in the form of protein and fat derived mainly from bacterial matter. Livesey (1990) calculated from published human balance studies that, on average, 30% of the energy of fermented carbohydrate in mixed human diets is converted to bacterial energy and lost to the feces. Thus, the efficiency of conversion of the energy of fermented carbohydrate to digestible energy is approximately 70%. From these studies, it was derived that the energy values of most NSP in mixed human diets can be estimated by multiplying the heat of combustion of NSP times the coefficient of their fermentability times 0.7 (i.e., the efficiency of conversion of fermented energy to digestible energy) (Livesey 1990). In agreement with the concept of Kleiber (1975), these energy values of NSP are partial digestible energy values, which are less than apparent digestible energy values because of the effect of dietary fiber on fecal protein and fat losses. However, for some fiber sources, especially whole grain cereal fibers, partial digestible energy values were lower than expected from the extent of NSP fermentation (Livesey 1990, Wisker et al. 1988 and 1996b). Therefore, energy values for these fibers cannot be estimated accurately from fermentation but should be determined in conventional digestible energy balance trials and can be calculated as proposed by Livesey (1989).

Partial digestible energy of NSP cannot be used completely for metabolism by the host, because there are energy losses in form of combustible gases and heat arising from fermentation and during production of ATP from the oxidation of SCFA, which is less efficient compared with ATP gains from the oxidation of glucose. Therefore, for the conversion of partial digestible energy of NSP to energy useful for humans or animals, i.e., net energy, partial digestible energy values have to be corrected for these losses (Livesey 1992).

At present, only a limited number of partial digestible energy values of dietary fiber have been determined experimentally in humans (Wisker and Feldheim 1990, Wisker et al. 1992, 1994 and 1996b). Because human balance studies are expensive and time consuming, such values are often measured in rats (Livesey et al. 1990 and 1995). Studies in rats are easier to perform, require less dietary material and cost less, and results can be obtained more rapidly than in humans. In addition, rat studies allow the investigation of NSP sources not yet licensed for human diets.

However, knowledge of how well the information obtained using rats applies to humans is limited. For several isolated fiber sources, a similar extent of NSP digestibility was observed in humans and rats (Nyman et al. 1986). On the basis of these results, the rat is considered to be a suitable model for humans for the estimation of energy values of NSP isolates (British Nutrition Foundation 1990, Livesey et al. 1995). However, it is not known whether this is also the case with NSP contained in mixed diets. Our group found that the apparent digestibility of NSP in mixed diets was lower in rats than in humans (Bach Knudsen et al. 1994, Wisker et al. 1996a). Therefore, the aim of the present study was to compare the energy values in humans and rats of various NSP sources contained in high fiber diets. The NSP sources studied were a mixture of fruits and vegetables, a citrus fiber concentrate, a barley fiber concentrate at two levels of protein intake, and coarse and fine whole meal rye bread. Partial digestible and net energy values of NSP were calculated on the basis of its fermentability and also on the basis of conventional digestible energy balances, independent of fermentation. Comparisons of the apparent digestibility of dietary energy, protein, fat and total NSP in humans and rats have been published recently (Bach Knudsen et al. 1994, Wisker et al. 1996a).

MATERIALS AND METHODS

Human experiments

Table 1. Designs of the human experiments [View Table]

Rat experiments

Table 2. Chemical composition of the experimental diets [View Table]

Chemical analyses

Calculations and statistical analysis

The partial digestible energy values of the additional NSP were calculated both from NSP fermentability and from conventional digestible energy balances. Partial digestible energy values of NSP were calculated from fermentability (DEV_ferm) by multiplication of the fermentability of NSP times the heat of combustion of NSP times 0.7 (efficiency of conversion of fermented energy to digestible energy) (Livesey 1990). Partial digestible energy values of NSP determined in digestible energy balances (DEV_{dig.energy balance}) were obtained by an equation that involves small magnification of measurement errors (Livesey 1989). This equation takes into account the heat of combustion of NSP, the intake of NSP and dry matter, and fecal energy losses during the balance periods (Livesey 1989): In this formula, DEV_{dig.energy balance} is the partial digestible energy value of the additional NSP (kJ/g), H_c,s is the heat of combustion of NSP, E_tf and E_cf are the gross energies of feces when the high fiber diets and the corresponding low fiber control diet, respectively, were eaten (kilojoules per balance period), M_td and M_cd are the dry masses of the basal portion of the high fiber diets and the corresponding control diet, respectively (grams per balance period), and M_s is the additional amount of NSP in the high fiber diets (grams per balance period).

The net energy value of NSP was calculated from fermentation as described by Livesey (1991a): NE_ferm = (1-A-B-C) × D × G × H. In this equation, NE_ferm is the net energy value of NSP, A (=0.3) is the apparent efficiency of converting the fermented carbohydrates to fecal energy (i.e., bacterial matter, unabsorbed SCFA and malabsorbed nutrients), B (=0.05) and C (=0.05) are the energy losses as gas and heat, respectively, during fermentation, D is the proportion of unavailable carbohydrate that is apparently digested, G (=0.85) is the gain of ATP by humans or rats per kilojoule of SCFA gross energy compared with that of glucose, and H is the heat of combustion of NSP.

Net energy values were also calculated from digestible energy balances as described by Livesey et al. (1995): Heats of combustion used for the calculations were as follows: cereal NSP, 17.6 kJ/g; fruit and vegetable NSP, 16.8 kJ/g; citrus NSP, 16.5 kJ/g (Livesey 1992).

The results within each dietary treatment group were examined by one-way ANOVA as outlined by Snedecor and Cochran (1973): where y_ij is the dependent variable, µ is the overall mean, _i is the effect of species (human or rat) and _ij is the normally distributed random residue. Species differences within each dietary treatment group were identified by Fisher's protected least significant difference test (Snedecor and Cochran 1973). The correlation between digestible energy values estimated from fermentation and digestible energy balances within species was examined by linear regression as described by Box et al. (1978): where y_i is the dependent variable,  is the intercept,  is the slope of the regression and _i is the normal distributed random residue. Differences were regarded as significant at P < 0.05. Ranks of the energy values estimated in humans and rats were tested by Spearman rank test (Snedecor and Cochran 1973). All statistical calculations were performed using general linear modeling (StatView software, Abacus Concepts, Berkeley, CA).

RESULTS

Table 3. Apparent digestibility of non-starch polysaccharides derived from different sources determined in humans and rats1 [View Table]

Fecal dry matter excretion after intake of NSP from fruits and vegetables, the citrus fiber concentrate and the barley fiber concentrate at high protein intake was significantly higher in rats than in humans, whereas it was similar in both species after intake of NSP from barley fiber at low protein intake and from coarse and fine whole meal rye bread (Table 4).

Table 4. Increase in fecal dry matter due to the consumption of non-starch polysaccharides (NSP) in humans and rats1 [View Table]

Correlations between partial digestible energy values calculated from fermentation (DEV_ferm) and values calculated from digestible energy balances (DEV_{dig.energy balance}) are shown in Figure 1. For both species, energy values estimated from fermentation (DEV_ferm) were higher than those calculated from energy balances (DEV_{dig.energy balance}) with the exception of the energy value of citrus NSP in humans, which was the same by both calculations. For most fiber sources, and particularly those with low energy values, the differences between energy values obtained by the two calculation procedures were greater in rats than in humans. This is seen from the regression intercept being significantly lower for rats than for humans.

Partial digestible energy values of NSP calculated from fermentation and from digestible energy balances are given in Table 5. In the human subjects, partial digestible energy values of NSP calculated from fermentation (DEV_ferm) were 4.4 to 11.4 kJ/g NSP. Consistent with the lower values in rats than in humans for fermentation of NSP from fruits and vegetables, citrus fiber concentrate, and barley fiber concentrate at high protein intake, the partial digestible energy values of these NSP were significantly lower in rats than in humans. For the partial digestible energy values of NSP from the other fiber sources there were no significant species differences.

Table 5. Partial digestible energy values of non-starch polysaccharides (NSP) calculated from NSP fermentability and from digestible energy balances in humans and rats1 [View Table]

When estimated from digestible energy balances, partial digestible energy values (DEV_{dig.energy balance}) of NSP from fruits and vegetables, the citrus fiber concentrate and the barley fiber concentrate at high and low protein intakes were also significantly lower (6.7 to 19.4 kJ/g NSP) in rats than in humans, whereas the values obtained for the other NSP sources did not differ significantly between humans and rats.

Net energy values calculated from fermentation and from energy balances are given in Table 6. These values are about 70% of the corresponding partial digestible energy values. From fermentability we estimated that NSP provided between 3.2 and 8.3 kJ net energy/g in humans and 1.8 and 5.5 kJ/g in rats. Species differences were significant for NSP in fruits and vegetables, the citrus fiber concentrate and the barley fiber concentrate at high protein intake. Net energy values of NSP calculated from digestible energy balances were 2.1 to 7.9 kJ/g in humans and -12.5 and 1.0 kJ/g in rats. Species differences were significant for NSP in fruits and vegetables, the citrus fiber concentrate and the barley fiber concentrate at high and low protein intakes.

Table 6. Net energy values of non-starch polysaccharides (NSP) calculated from NSP fermentability and from digestible energy balances in humans and rats1 [View Table]

The Spearman rank correlation for digestible energy values for humans and rats was 0.414 when calculated from NSP fermentation and 0.143 when calculated from digestible energy balances. In none of the cases was the rank correlation significant.

DISCUSSION

Under these experimental conditions, the fermentation of the additional NSP consumed during the high fiber diet periods had to be calculated for both human subjects and rats from differences in NSP intake and excretion during high fiber and corresponding low fiber diet periods. Using these calculations, fermentation of NSP in the basal diet has been taken into account. Such calculations are valid only under conditions in which there is no effect of the additionally consumed NSP on the fermentation of NSP contained in the basal diet. Key and Mathers (1993b) calculated by multiple linear regression whether beans as additional fiber sources for rats altered the fermentation of NSP from whole meal bread in the basal diet. The proportion of whole meal bread was kept constant whereas the contribution of beans increased from 0 to 450 g/kg at the expense of sucrose and casein, leading to an increase in the dietary NSP content of from 27.9 to 96.6 g/kg. The presence of additional bean NSP had no effect on the fermentation of NSP in bread. A similar study in pigs also showed that additional pea NSP did not affect the fermentation of wheat NSP contained in the basal diet (Goodlad and Mathers 1991). Therefore, we assume that in the present study the fermentation of NSP in the basal diets was also not affected by the additional NSP. This assumption is supported by the fact that the basal diets contained highly fermentable NSP (Bach Knudsen et al. 1994, Wisker et al. 1996a).

The main factor influencing the availability of the energy provided by fiber is fermentability. The additional NSP consumed during the high fiber diet periods was in most cases less fermented in rats than in the human subjects with the exception of NSP from coarse whole meal rye bread, which was degraded to a somewhat greater extent in rats. These results are consistent with our previous findings that the apparent digestibility of total NSP in nine of the 10 experimental diets was lower in rats than in humans (Bach Knudsen et al. 1994, Wisker et al. 1996a). In these studies, differences between species were greater with the high fiber diets than the low fiber diets with the exception of the high fiber diet containing coarse whole meal rye bread. In addition to probable species differences in the capacity of large intestinal bacteria to ferment certain fiber sources (Bach Knudsen et al. 1994), the lower digestibility of the additional dietary NSP in rats compared with humans may have been due to species differences in colonic transit time, which influences the degree of fermentation (Van Soest et al. 1982). Several studies have shown that transit times are shorter in rats than in humans (Cummings et al. 1976, Raczynski et al. 1982), leaving less time for bacterial NSP degradation in rats. Although in both species transit time can be shortened by an elevated fiber intake (Cummings et al. 1976, Raczynski et al. 1982), the magnitude of this effect may differ between species, with consequences for the extent of fermentation.

Differences and similarities between human subjects and rats may have been influenced by the fact that young female subjects have been compared with young male rats. In general, there is large individual variation in fecal weight and in the fermentation of NSP not only in rather homogenous groups of human subjects (Bach Knudsen et al. 1994, Southgate and Durnin 1970, Wisker et al. 1996b) but also in rats (Bach Knudsen et al. 1994, Wisker et al. 1996a). In their study on energy conversion factors, Southgate and Durnin (1970) found no significant influence of sex on the digestibility of energy and the fermentation of pentosans and cellulose, but cellulose was significantly better digested in elderly men than in young men, whereas there were no significant differences between young and elderly women. Stephen et al. (1986) studied in a group of 19 men and 11 women (age 17 to 62 y) the effect of increasing quantities of wheat fiber on stool weight, transit time and other fecal variables. For each dose of wheat fiber, seven or eight subjects were studied. When they consumed the basal diets, men had higher fecal wet weights than women due to shorter intestinal transit times, but there were no age-related differences. However, these authors provided no information on whether NSP fermentation was lower and fecal energy losses higher in the men than in the women, which would lead to lower energy values of NSP in men. For rats, no data are available on sex- and age-related influences on fecal output and NSP fermentation. However, Weaver et al. (1992) found significant strain differences in the production of SCFA during cornstarch fermentation in vitro when rat feces were used as inoculum. Whether our results would have differed if other groups of rats or human subjects (e.g., men, mixed groups of men and women, older persons) had been studied is unknown. Non-starch polysaccharide fermentation and fecal output in the young, nonconstipated women participating in our study may be between those of young men and older persons. This is supported by the fact that the increases in fecal weight due to the single fiber sources (1.8 to 6.1 g/g NSP; Wisker and Feldheim 1990, Wisker et al. 1992, 1994 and 1996b) were in the range of values found in other studies in which comparable fiber sources were investigated and modern methods of fiber analysis were used (Cummings 1986).

In the present study, the human subjects consumed close to their maintenance levels, whereas the energy intake of the rats was approximately two times the requirement for maintenance (Zhao et al. 1995). The higher energy intake in rats could cause an overflow of nutrients to the large intestine, which potentially could influence the fermentation of dietary fiber; however, it is unlikely that the feeding levels used for rats in the present study had a major effect on the fermentation of dietary fiber components. In a study with growing rats fed from 5 to 13 g/d (corresponding to approximately one to three times the maintenance level), the feeding level did not influence the digestibility of soluble and insoluble dietary fibers (Larsen et al. 1991). The digestibility of dry matter, however, was significantly and negatively affected by the daily food intake level.

Consistent with the significantly lower fermentation of NSP in rats than in humans for three of the six fiber sources studied, significantly lower partial digestible energy values were calculated from fermentation (DEV_ferm) for these NSP in the rat experiments compared with the human studies. The ranking of the energy values differed between the species. In the human studies, the highest and lowest values were found for citrus NSP and barley NSP at the low protein intake, respectively, whereas in the rat experiments the highest energy value was calculated for NSP in coarse whole meal rye bread and the lowest for barley fiber at the high protein intake.

Partial digestible energy values of NSP on the basis of its fermentability (DEV_ferm) were obtained under the assumption that 30% of the fermented energy is lost to feces in the forms of bacterial mass, unabsorbed SCFA and (probably) unabsorbed dietary nutrients. Mean fecal energy losses due to fermentation in such a quantity have been calculated for NSP in mixed human diets (Livesey 1990). Losses between 20% (Livesey et al. 1990) and 40% (Davies et al. 1991) have been found in rats fed different sources of unavailable carbohydrates, so that an average conversion of fermented energy to fecal energy of 0.3 is also regarded as suitable for rats (Livesey 1991a).

In humans, however, it has been shown that energy values of NSP cannot always be estimated from the extent of fermentation. This is especially the case for NSP derived from whole meal bread and other whole grain cereal foods, which were found to lead to higher energy losses than could be expected from fermentation (Livesey 1990, Wisker et al. 1988 and 1996b). Thus, energy values of NSP estimated from energy balances by using dry matter intake and fecal energy losses during consumption of a high fiber diet and a corresponding low fiber diet will be more accurate. This is also supported by the finding that dry matter intake and fecal energy losses can be determined with higher precision than can NSP fermentability (Livesey 1989, Livesey et al. 1995).

With the exception of the value for citrus fiber in humans, partial digestible energy values calculated from energy balances (DEV_{dig.energy balance}) were lower than those estimated from NSP fermentability (DEV_ferm) in both human subjects and rats. As expected, in humans these differences were greatest for coarse whole meal bread (Livesey 1990, Wisker et al. 1988). However, for NSP from fruits and vegetables, citrus fiber, and barley fiber at high or low protein intake, the differences between these energy values were higher in rats than in humans. This is consistent with the greater increase in fecal dry matter losses in rats than in humans during the ingestion of these fiber sources being mainly due to a higher excretion of unfermented dietary fiber and of residues (mostly Klason lignin) in rats than in humans (Bach Knudsen et al. 1994). The differences between humans and rats in fecal dry matter excretion are most likely related to shorter transit times in the latter. Shorter transit times lead to a lower degree of NSP fermentation (Van Soest et al. 1982) and also give a higher yield of microbial cells per gram of fermented carbohydrate because of a diminished requirement of bacteria for maintenance energy (Cummings 1984).

It is unlikely that the lower energy values estimated in rats compared with humans are caused by a systematically incomplete recovery of fecal materials in rats fed the low dietary fiber control diets. In two studies using rats, total recoveries of markers in feces up to 114 h after dosing were estimated to be 96% (range 90-101%) (Bach Knudsen et al. 1991) and 100% (range 96-106%) (Hansen et al. 1992). Although in both studies the lowest recovery was obtained with low dietary fiber diets, the results from these studies did not suggest any systematic effect of dietary fiber level on the fecal recovery.

With respect to the results obtained in rats, our findings differ from those of Livesey et al. (1995). These authors found a good agreement in rats in energy values calculated either from NSP fermentation or from conventional digestible energy balances when several starch-free or low starch fiber supplements were added to fiber-free semipurified diets. The fiber concentrates used in our studies (citrus fiber, barley fiber) also contained no starch or only very small amounts of starch, but with these fiber sources we observed the biggest differences between humans and rats. Thus, the starch content of the additional NSP cannot have been the reason for these discrepancies.

In rats, does it make a difference whether additional fiber sources are added to fiber-free semipurified diets or to mixed diets as consumed by humans? Nyman et al. (1986) found a good agreement between humans and rats in the fermentation of dietary fiber isolates. In the human studies, these isolates were added to mixed diets providing about 20 g dietary fiber/d (i.e., 37 g dietary fiber/kg diet dry matter), whereas they were given to rats in addition to semipurified diets. In our studies, foods used for the rat diets had been freeze-dried, which may have increased the amount of resistant starch (Englyst et al. 1982). We did not measure resistant starch separately, but some resistant starch was likely determined together with NSP (Bach Knudsen et al. 1988), because the analytical procedure used for NSP determination did not have a solubilization step for resistant starch. Therefore, theoretically more fermentable material could have reached the large intestine in rats than in humans. A higher substrate supply stimulates bacterial growth and thus could have led to higher energy losses in the form of bacterial protein in rats than in humans. However, there are two objections to these assumptions: 1) Similar quantities and types of starch were present in both the high fiber diets and the corresponding low fiber diets. Therefore, any resistant starch formation during freeze-drying and any possible effects of resistant starch on fecal energy losses should have been similar for high fiber and corresponding low fiber diets and thus have been taken into account by the calculations used for NSP fermentation and energy values. 2) A more intensive fermentation of dietary residues stimulates the synthesis of bacterial mass, which is followed by a higher excretion of microbial nitrogen by the feces (Mason and Palmer 1973). However, calculated per gram of additional NSP ingested, fecal protein losses tended to be higher in humans than in rats, except for barley fiber at high protein intake. This finding is consistent with the higher fermentation of NSP in humans than in rats (Bach Knudsen et al. 1994, Wisker et al. 1996a).

Physico-chemical factors such as particle size of cereals may affect fecal energy losses in humans, probably because large particles can protect starch from small intestinal digestion (Livesey 1990, Wisker et al. 1988). However, particle size seems not to be important in rats, probably because rats finely chew large particles in their diets and because smaller particles of digesta are formed in the stomach and small intestine of rats than of humans (Livesey 1991b). Consistent with these findings, rats in our studies digested NSP in coarse whole meal rye bread better than NSP in the fine bread.

There are anatomical differences between humans and rats with respect to the digestion of fat; rats have no gallbladder, cannot concentrate bile and have a nearly constant flow of diluted bile into the duodenum (Björnhag 1992). Rats may therefore be unable to emulsify large meals of fat sufficiently, which potentially may lead to higher losses of fat and thereby energy to the feces. Despite these differences and the relatively high dietary fat levels (131-195 g/kg) compared with standard rat diets (~30 g/kg) (Eggum 1973), however, there was an excellent agreement in the digestibility of fat between humans and rats. Moreover, calculated per gram of additional NSP, fecal losses of fat were very similar in humans and rats (Bach Knudsen et al. 1994, Wisker et al. 1996a). In addition, there seems to be no effect of fat content of rat diets on NSP digestibility. Key and Mathers (1993a) showed that increasing maize oil from 30 to 170 g/kg diet had no influence on NSP digestibility in rats fed either white or whole meal bread. In rats, therefore, effects of a fat intake as used in the present study on energy values of NSP are rather unlikely.

In conclusion, our results suggest that the rat is not always a suitable model for the prediction of energy values of NSP in mixed diets consumed by humans, mainly because of differences in NSP fermentation between humans and rats. In addition, when partial digestible energy values of NSP measured in digestible energy balance trials are regarded as standard, energy values of some types of NSP contained in mixed diets could not be calculated accurately from NSP fermentability. Whether fermentation of NSP and fecal energy losses in rats are affected by the type of the basal diet remains to be studied.