The carbohydrate fraction of dietary fiber is a heterogeneous and complex mixture of different combinations and linkages of monosaccharides that can best be viewed as a biological entity rather than a chemically defined component of the diet (Van Soest 1994). Given the amorphous nature of fiber and the methodologies available for its quantification, its definition is a function of the particular analytical methods used. The recognition that the quantification of crude fiber does not adequately measure many important components of the plant cell wall has led to the development of many new methods of fiber analysis. Some methods such as neutral detergent fiber (NDF)³ measure one portion (the insoluble fiber) of total dietary fiber, yet there are well-defined relationships between NDF intake and physiological actions such as digestibility, mineral availability and other nutrient interactions. Total dietary fiber (TDF) (Prosky et al. 1985) is a recent analytical method recognized as an official method of the Association of Official Analytical Chemists (AOAC), and this methodology has taken on an important role in food labeling and human nutrition. There have been many reports devoted to the analytical aspects of this method, but there are few data relating TDF content of foods to physiological action. The interaction of fiber with protein and fat affects the digestibility of these nutrients and, consequently, the metabolizable energy (ME) content of the diet. The effects of NDF, as well as some other types of fiber (e.g., nonstarch polysaccharides) have been established. However, in vivo digestibility of TDF in mixed diets has not been reported.

Data on the interactions among energy-yielding nutrients are critical for estimating the ME value of foods and diets for experimental, clinical and regulatory purposes. The U.S. Nutrition Labeling and Education Act (NLEA) of 1990 requires that the label for food intended for human consumption bear nutritional information, including the energy (number of calories) in each serving or other unit of measure of the food. The most appropriate method for determining the energy in a serving of a food is unclear. Caloric value may not simply be additive ME from fat, protein and carbohydrate provided by factorial equations but may be a function of the interaction of these nutrients with dietary fiber. Thus, factorial equations may be inadequate. However, factorial equations are the only equations approved for use in food labeling (Federal Register 1993).

In published studies, there is a large range in predicted ME content of diets (Livesey 1990), and the range may be related to different fiber sources and methodologies used to determine fiber intake in these studies. There are few data concerning measured and predicted ME content of diets varying in both fiber and fat concentration and on the specific effect of TDF on energy availability. Nutrient interactions and the associated inaccuracy of predicting ME content of diets may contribute to food labeling inaccuracies resulting in potential deleterious consequences for consumers.

Historically, the factorial approach has been used to calculate the energy content of diets (Atwater and Bryant 1900), and the ME content of dietary fiber has been assigned energy values between 20.92 and 12.97 kJ/g (reviewed in Livesey 1990). One impediment to determining the ME value of fiber, particularly for use in factorial prediction equations, is the lack of data on in vivo fiber digestibility in humans. Furthermore, much of the data on fiber digestibility is based on NDF, the Southgate method (nonstarch polysaccharides) or the Uppsala TDF method, and not the Prosky TDF method. Previous published data on in vivo TDF digestibility in humans are limited to four diets that contained chemically defined fiber sources (soy polysaccharides, oat fiber, carboxymethylcellulose and gum arabic) used in nutrient supplements (Sunvold et al. 1995). There are no data on TDF digestibility from mixed fiber sources of typical American diets.

In addition to factorial equations, several empirical formulas were published (Levy et al. 1958, Miller and Judd 1984, Miller and Payne 1959, Southgate 1975) and reviewed (Livesey 1990). These empirical formulas have a bias ranging from 12% to +6%, which increases with increasing dietary fiber intake. On the basis of these findings, Livesey (1991) developed a new empirical formula derived from published data from 43 human diets with 1.9-93.1 g/d of dietary fiber from various sources and based on fiber values from several different methodologies (Southgate, Asp, NDF and Uppsala). This formula was developed and tested using the same data sets; it has not been verified with either an independent data set or a set of data in which TDF intake was measured exclusively.

The Prosky TDF method has become important in human nutrition and in the food industry. The present study was conducted with the following objectives: 1) to measure and compare the digestibility of TDF, NDF and "soluble" fiber (TDF minus NDF) of typical Western mixed diets containing different amounts of fat and fiber; 2) to compare the effect of TDF, NDF and "soluble" fiber on energy availability; and 3) to simultaneously measure the ME content of complete diets that varied in both fat and fiber content and compare determined and predicted ME values using Livesey's (1991) equation.

SUBJECTS AND METHODS

Physical characteristics of the 15 healthy, free-living subjects are presented in Table 1. Before acceptance into the study, a medical screening, including blood and urine profiles, was performed on each subject. Physiological variables were in the normal range for all subjects. Five or six subjects were randomly assigned to a set of three diets that were formulated to contain one of three levels of dietary fat (low: 15% of energy from fat; medium: 30% of energy from fat; or high: 45% of energy from fat). Each set of three diets, within each level of fat, contained different levels of dietary fiber (low, medium or high fiber), so that each subject consumed three diets with different levels of dietary fiber and one level of dietary fat. The terms high, medium, and low refer only to the relative amounts of fat and fiber in the nine diets. Two subjects (#9 and #11) agreed to repeat the study protocol. After completing one collection protocol, these two subjects were assigned to a different level of fat, and observations from these two subjects at each level of fat were considered independently. Thus, there were a total of 53 observations (five or six subjects per fat level, three levels of fat, and three fiber levels per fat level).

Table 1. Physical characteristics of subjects [View Table]

Subjects were fed three meals daily and all meals were prepared at the Beltsville (MD) Human Nutrition Research Center Human Studies Facility. Diets were composed of foods typically found in a Western diet. Menus for six of the nine diets are presented in Table 2. Nutrient composition of the diets was calculated from USDA Handbook No. 8 values (USDA 1976-1989; Watt and Merrill 1963). The specific amount of food consumed was based on the maintenance energy requirement estimated from one of two methods. For some subjects, the maintenance energy requirement was based on prior measurements of 24-h energy expenditure in a room calorimeter plus 15% (Seale et al. 1991). For other subjects, the maintenance energy requirement was calculated based on age, weight, height and gender (Harris and Benedict 1919). Based on the measured (or estimated) maintenance energy requirement, the amount of daily food intake was calculated. Subjects were fed the same items, and the same proportions (based on individual maintenance energy requirements) of each item, each day for 14 d for each combination of dietary fat and fiber. Therefore, the ratio of all nutrients was constant for all subjects for a given diet. Subjects were weighed daily to verify weight maintenance throughout the study; this was achieved for all subjects during the 14-d periods. Mean energy intake was similar for each fat level. After a 9-d adaptation period to the diet, the subjects were given brilliant blue (Warner and Kenkinson, St. Louis, MO, ~20 mg) in a gelatin capsule to consume with their evening meal, and they were instructed to collect all feces and urine voided over the next 5-d period. At the end of the 5-d collection period, another dose of brilliant blue was administered. There was a 2-wk washout period before subjects began adaptation to the next level of dietary fiber. Food intake and composition were not monitored during the washout period.

Table 2. Menu of six diets, 24-h intake [View Table]

During the collection period, total urine was collected, weighed daily, sampled (10% of daily production), and the daily subsamples were pooled. Total fecal output was collected, and feces were frozen, pooled, weighed and homogenized with water and ice in a blender before being freeze-dried. Diets were prepared for chemical analysis by homogenizing the food in a blender with ice and water before being freeze-dried. Diets, feces and urine were analyzed for combustible energy by adiabatic bomb calorimetry (Parr Instrument, Moline, IL) and for nitrogen by combustion (Leco, St. Joseph, MI). Diets and feces also were analyzed for fat (methylene chloride extraction, CEM, Matthew, NC), ash (muffle furnace), TDF (Prosky 1985) and NDF (Robertson and Van Soest 1981). Before being analyzed for NDF, diets and feces were incubated with heat-stable -amylase for 1 h (85°C) and then with protease for 1 h (85°C). Total carbohydrate (dry matter basis) was calculated as Soluble fiber in the diet and feces was calculated as

The ME value of each diet was calculated as the difference between intake energy (total combustible energy of the diet) and the sum of combustible energy excreted in feces and urine. Predicted ME was calculated using the formula of Livesey (1991): where E is combustible energy intake, U is unavailable complex carbohydrate intake, and N is nitrogen intake. Total dietary fiber, NDF, soluble fiber, protein and fat digestibility were calculated as the difference between nutrient intake and fecal loss, expressed as a percentage of nutrient intake. Therefore, in this paper, all references to digestibility refer to apparent digestibility.

Data were analyzed as a crossover design study using analysis of covariance (ANCOVA; SAS version 6.08, SAS Institute, Cary, NC). The statistical model included terms for fat, subject within fat level, fiber, and the fat × fiber interaction. The subject-within-fat-level term was used to determine least-squares means for the effect of fat. The residual error was used to determine least-squares means for fiber and the fat × fiber interaction. Before ANCOVA, variables were tested for homogeneity of variance (by plotting residuals and predicted values and by the degree of Spearman correlation coefficient between residuals and predicted values) and for normality. Estimates of the slope and intercept were determined from the ANCOVA model statement. Generalized linear models were used to calculate regression parameters between ME and TDF intake.

RESULTS

Table 3. Chemical composition of nine diets1 [View Table]

Calculated crude fiber and measured TDF, NDF and soluble fiber concentrations are presented in Table 4. Diets were formulated on the basis of published crude fiber values to contain ~2.0, 1.0 and 0.5 g/100 g crude fiber for the high, medium and low fiber diets, respectively (Table 4). Crude fiber data were used as the measure of fiber for formulation purposes because values are available for the most common foods. Total dietary fiber values were 3-9 times greater than the crude fiber values, and the difference between TDF and crude fiber concentration appeared to be greatest for the three low fiber diets. As expected, TDF concentration was ~1.5-3 times greater than NDF concentration. The calculated soluble fiber concentration ranged from 1.4 to 2.9 g/100 g dry matter, and there was no consistent relationship between TDF or NDF and the soluble fiber concentration of these diets.

Table 4. Fiber fraction composition of nine diets [View Table]

Least-squares mean daily nutrient and energy intakes are presented in Table 5. Daily fat intake ranged from 38 to 164 g. Fat intakes were markedly different for the three different fat levels. Fat intake (expressed as percentage of total energy) from the high and medium fat diets was similar to the current reported average fat intake of the population in the United States (Department of Health and Human Services 1990). Fat intake from the low fat diets was consistent with current recommendations for fat intake.

Table 5. Daily nutrient and metabolizable energy intakes and fecal and urinary energy losses of humans fed diets of different fat and fiber content1,2 [View Table]

Daily TDF intakes ranged from 15 to 46 g. In the United States, mean fiber intake is 18 g/d for men and 12 g/d for women, and similar intakes were achieved when subjects consumed the low fiber diets. Fiber intake from the high and medium fiber diets approached or exceeded current recommendations for a healthy diet, thus providing excellent fiber intakes with typical dietary items. Daily NDF intakes followed a similar pattern but were generally lower than TDF intake. Soluble fiber intakes were lower than NDF intake and did not appear to follow any particular pattern. Daily protein intake ranged from 94 to 130 g, and total carbohydrate intake ranged from 251 to 421 g. Fecal energy content was higher when subjects were fed the high fat diets than the other diets, but urinary energy output was similar across most treatments.

Apparent nutrient digestibility data for the nine diets are presented in Table 6. As a consequence of significant fat and fiber interactions, it was possible to examine the effect of fiber intake on apparent nutrient digestibility within each fat level only. Effects of fiber intake on nutrient digestibility were similar for each of the three fat levels. In general, ME, fat and protein digestibility were lower (P < 0.05) when subjects consumed the high fiber diets compared with the low fiber diets. However, when subjects consumed the low fat diets, there was no effect of fiber intake on fat digestibility. Furthermore, there was little consistent effect of fiber intake on TDF, NDF or soluble fiber digestibility.

Table 6. Apparent digestibility of fat, fiber fraction and protein and metabolizable energy coefficient of humans fed diets of different fat and fiber content1 [View Table]

Total dietary fiber digestibility (across all diets) ranged from 67 to 82 g/100 g, NDF digestibility ranged from 48 to 81 g/100 g, and soluble fiber digestibility ranged from 59 to 97 g/100 g. Total dietary fiber digestibility (paired t test) was higher (P < 0.02) than NDF digestibility (grand means of 73 and 64 g/100 g, respectively). The grand mean for soluble fiber digestibility was 82 g/100 g.

Predicted daily ME intake was calculated and compared using both TDF and NDF as the unavailable carbohydrate sources (U). The measured ME intake and the predicted ME intake were not different when either TDF or NDF intake were used as U. Mean predicted ME intake (10.48 MJ using TDF intake and 10.38 MJ using NDF intake) and measured ME intake (10.42 MJ) did not differ (P > 0.05), and the regression slope between actual and predicted ME intake was not different from zero (Fig. 1). For comparison, using Atwater's factorial equation to predict ME intake overestimates (P < 0.001) the measured ME intake (10.89 MJ vs. 10.42 MJ).

The linear relationships between ME intake (kJ/g dry matter) and TDF, NDF and soluble fiber intake (g/g dry matter) are presented in Figure 2. Each point represents one observation from each subject. Many points overlap as a consequence of the controlled fiber intake and the similarity in ME intake for the different subjects. The slope of the line represents the change in ME of the diet as a function of a change in fiber intake (ME/g fiber consumed). For TDF and NDF, the slopes were similar for both types of fiber and for all three fat levels (22 kJ/g TDF and 21 kJ/g NDF for the high fat diets, 34 kJ/g TDF and 31 kJ/g NDF for the medium fat diets, and 31 kJ/g TDF and 41 kJ/g NDF for the low fat diets). For soluble fiber, the slopes for the high fat and medium fat diets were positive (92 and 52 kJ/g soluble fiber, respectively) and the slope for the low fat diet was negative (77 kJ/g soluble fiber).

DISCUSSION

The relationship between fiber intake and fat and protein digestibility of diets fed in this study was consistent with previously published data (Calloway and Kretsch 1978, Farrell et al. 1978, Kelsay et al. 1978, Southgate and Durnin 1970). As the fiber content of the diet increased, fat and protein digestibility decreased. There was little difference in fat digestibility across fiber levels for the low fat diets, perhaps because of the similarity of fiber intake for the low fat, medium fiber diet and the low fat, low fiber diet or because of the limited fat intake from these diets.

The digestibility of TDF was generally greater than that of NDF. This finding is likely a consequence of the high digestibility of soluble fiber, even though soluble fiber intake typically accounted for less than one-half of the TDF intake. The least-squares mean digestibility of NDF in these mixed diets (grand mean = 61%) is consistent with the NDF digestibility previously reported for diets containing different sources of fiber, including Solka floc (10-24 g NDF/d, Slavin et al. 1981), wheat bran (9-21 g NDF/d, Marlett and Johnson 1985; 33-54 g NDF/d, Farrell et al. 1978), and fiber from fruits and vegetables (2-26 g NDF/d, Kelsay et al. 1981a and 1981b).

There was little consistent effect of different diets on fiber digestibility (TDF, NDF and soluble), and fiber digestibility was more variable than the digestibility of the other measured dietary components. In several previously reported human studies, fiber digestibility (NDF) decreased (Farrell et al. 1978, Kelsay et al. 1981a, Slavin et al. 1981) or did not change (Kelsay et al. 1981b) in response to changes in fiber intake. In one study using wheat bran as the fiber source (Marlett and Johnson, 1985), NDF digestibility increased with an increase in fiber intake. The inconsistency of the reported effects of fiber intake on fiber digestibility may be a result of the use of different fiber sources (wheat bran, Solka floc, fruits and vegetables) and improvements in NDF analysis.

The large individual variability in fiber digestibility may result from individual differences in anaerobic microbial species composition and population size, and fermentation rate. Large individual differences in fiber digestibility have been reported (Slavin et al. 1981, Southgate and Durnin 1970). Moreover, the large individual variability in TDF and NDF digestibility suggests that it is difficult to determine precisely a reliable value for use as the ME value of dietary fiber. The ME value of fiber is a function of combustible energy content and digestibility (Livesey 1990). Because the combustible energy of fiber is relatively consistent, changes in digestibility will affect the ME value. Considering the ranges in fiber intake and digestibilities measured in this study, the calculated ME value (product of weight of fiber digested, combustible energy content and availability of volatile fatty acids) for fiber [assuming gross energy value of 17.2 kJ/g and 70% availability of volatile fatty acids (Livesey 1990)] ranges from 2.8 to 11.2 kJ/g fiber consumed (25% CV).

Based on data from the present study, ME intake can be predicted accurately using Livesey's equation, and both TDF and NDF intake can be used as the measure for U. Livesey's data set for predicting the ME of mixed diets relied on dietary fiber analyzed by several methods. No systematic bias was associated with the different analytical methods except for a possible underestimation of non-starch polysaccharides with the NDF method (Livesey 1991). The data from the present study suggest that the underestimation of U from NDF analysis is insignificant. In the present study, when NDF was used as the measure of U, the predicted ME intake was not different from the measured ME intake. Neutral detergent fiber methodology may underestimate non-starch polysaccharide intake (Livesey 1991); however, NDF represents the majority of TDF of these diets and, thus, there is little difference between predicted ME using either TDF or NDF.

The relationship between dietary fiber and ME concentration was similar for both TDF and NDF and was independent of the amount of dietary fat. As TDF and NDF dietary concentrations increased, ME density decreased (slope <1). This decrease is probably a consequence of the decrease in digestibility of fat and protein associated with the increased dietary fiber concentration and with the replacement of starch with fiber. On the other hand, the effect of the concentration of soluble fiber in the diet appeared to be inconsistent. For the high fat diet, there was a higher dietary ME density associated with a higher soluble fiber concentration. For the medium fat diet, there was no change (slope = 0) in ME density associated with dietary soluble fiber concentration, and for the low fat diet, ME density decreased as soluble fiber concentration increased (slope >1). These differences may be associated with the fat content of the diet or with other dietary components. Dietary TDF and NDF concentrations appeared to have similar effects on ME density of these diets. However, the effects of soluble fiber concentration appeared to be different than the effects of TDF and NDF.

As a consequence of the decrease in digestibility of fat and protein, the actual ME value of mixed diets depends on the overall composition of the diet and, thus, it may be more difficult to predict the ME content of mixed foods based on the amount of the individual macronutrients. Given the wide range in fiber digestibility and the nature of the interaction between fiber and other dietary components, it may be difficult to assign a single ME value for fiber. Presumably, the ME value will depend on the type of dietary fiber as well as the overall composition of the diet. The effect on ME of increasing the fiber content of these diets may result from the replacement of starch (which is generally readily digested) with fiber (which is more resistant to digestion) as well as from the effect of fiber on the digestion of other nutrients. From the data obtained from the mixed diets fed in this study, it is not clear if the interaction between fiber and fat is consistent for different sources of fiber and fat.

Interactions among fiber, protein and fat affect the digestibility of nutrients and ME available from mixed diets. Based on regression of TDF or NDF intake on ME density, the fiber measured by these two analytical methods appears to have a similar effect on ME availability from a mixed diet. The slope of these regression lines (30 kJ/g fiber consumed) indicates that if fiber intake from typical American diets were doubled (without a concomitant increase in total energy intake) from 18 g/d (for men) to 36 g/d, there would be a decrease of ~540 kJ in daily ME intake. Clearly, fiber from some sources may be fermented and the volatile fatty acids may contribute to the energy available from the diet. This value does not represent the ME value of fiber per se. The overall effect of increasing mixed fiber sources, particularly those that contain a substantial amount of insoluble fiber, is to decrease the digestibility of energy-yielding nutrients and to decrease the amount of ME available from the diet.

Increasing daily fiber intake may have many benefits and may reduce risk factors associated with certain diseases. One challenge in dietary fiber research is to develop an analytical system that adequately measures the fiber content of foods as related to demonstrable physiological actions. In the mixed diets fed to these subjects, TDF digestibility was typically higher than NDF digestibility, and increasing TDF and NDF intake decreased the ME density of these diets. The decrease in ME density is possibly associated with both the limited energy availability from plant fiber carbohydrates and the inhibitory effect of fiber on fat and protein digestibility. When used to predict ME content of these mixed diets, TDF and NDF intake provided accurate and precise predictions. The similarity in prediction is likely a consequence of the relatively high proportion of insoluble fiber in these typical American diets.