 |
INTRODUCTION |
High fiber, low fat and low energy diets are recommended to people attempting to lose weight, but there are no direct determinations of the energy values of these diets. For such diets, there remains a major uncertainty about the accuracy of food energy assessment systems, which range from the early Atwater and Bryant (1900)
to the most recent systems in the U.S. (FDA 1993) and Europe (European Council Directive 1990). This lack of information was noted previously by a Life Sciences Research Office (1983) expert panel on the Atwater systems of food energy assessment.
The requirements for accurate food energy evaluation systems have been recorded previously (Baer et al. 1997, Livesey 1990, Merrill and Watt 1973), but not in relation to weight-reducing diets. The interrelation among energy intake, obesity and disease incidence (Livesey 1995a, Royal College of Physicians 1983) makes it necessary that food energy assessment systems be evaluated under the condition of negative energy balance or weight loss in addition to maintenance conditions for energy balance and body weight. The lack of information on high fiber submaintenance diets arises in part because food energy evaluation systems were designed for application in the normative maintenance situation (Livesey 1995a, Merrill and Watt 1973). We therefore undertook a study on energy availability from a high NSP, low fat, weight-reducing diet, comprised of foods similar to, but not identical with, one that has been popularized (Eyton 1982). Our principal question is the following: do any of the food energy assessment systems predict the availability of energy from a high fiber submaintenance diet? This was answered by comparing the determined energy availability from such a weight-reducing diet with that predicted by variously recommended food energy assessment systems (European Council Directive 1990, FDA 1993, Livesey 1990 and 1991a, Merrill and Watt 1973).
A second question we address is the reliability of food energy assessment systems for maintenance diets that contain variable amounts of dietary fiber. This has been studied previously (Baer et al. 1997, Barry et al. 1995, Behall and Howe, 1995, Livesey, 1990, 1991a, 1992, 1993, 1995a and 1995b, Miller and Judd 1983, Southgate and Durnin 1970, Wisker et al. 1993, 1996a and 1996b), but certain important questions remain. One is the reliability of the Atwater specific system of food energy assessment (Merrill and Watt 1973), which is held in high regard in the U.S. as the gold standard. When the specific factor system was originally tested, it was found to be accurate to within ~2%. It has also been suggested that it is without bias with increasing dietary fiber intake from conventional foods (Livesey 1990, Mathews 1995). The performance of the Atwater specific systems against other systems that are designed to take account of dietary fiber is one of the questions asked here.
Importantly, a modified factorial system of food energy assessment has been proposed to take account of the variable dietary fiber content of diets, in which the proposed conventional metabolizable energy factor for dietary fiber is 8.4 kJ/g (Livesey 1990). Until now, there has been no test of its validity on the basis of a data set obtained independently of that on which the proposal was formulated. For the first time, this study appraises the recent food energy general formula, based on gross energy, as introduced by the FDA (1993). Furthermore, a gross energy-based formula proposed by Livesey (1991a), which had been developed from information on 43 diets of varied dietary fiber content, has recently been independently validated on diets of variable fat and fiber content (Baer et al. 1997). We examine here whether such independent validation extends into another laboratory using a different method of dietary fiber analysis (see below).
Other systems assessed include the Atwater general factor system (Atwater and Bryant 1990, Merrill and Watt 1973), which is permitted in U.S. Food Labeling Regulations (FDA 1993), and a European general factor system, which uses estimates of available carbohydrate by difference and attributes no energy value to dietary fiber (European Council Directive 1990). These two systems make no provision for the influence of variation in dietary fiber on energy availability and are thus expected to show bias with high fiber diets (Baer et al 1997, Livesey 1990, Merrill and Watt 1973), but they were included here because they remain within food labeling provisions. However, we have excluded other food energy assessment systems that have previously been found to display bias when applied to high fiber diets (Livesey 1990), namely, those of Miller and Judd (1983), Levy et al. (1958), Southgate (1975) and the British system (Paul and Southgate 1978).
Our measure of dietary fiber for these assessments has been what has been called unavailable (complex) carbohydrate; this is the sum of non-starch polysaccharides (NSP)4 and resistant starch (RS) (British Nutrition Foundation 1990). This approach has not been used in previous energy assessment systems and, in principle, is essentially equivalent to total dietary fiber. For the present purposes we formulated two mixed, maintenance diets of differing NSP content but of similar protein, fat and available carbohydrate content. In addition, a third maintenance diet was formulated to contain large amounts of carbohydrate from barley, a food that has yielded varied results on energy availability in human feeding trials (Merrill and Watt, 1973, Miller and Judd 1983). Finally, the high fiber, low fat, low energy diet mentioned above was included in our investigation.
| |
SUBJECTS AND METHODS |
Approval of ethics.
The studies were approved by the Institute of Food Research Human Research Ethics Committee and the Dunn Clinical Nutritional Centre Ethics Committee. All volunteers gave informed written consent.
Foods.
All foods used in this study are itemized in the Appendix. They were purchased from local supermarkets with the exceptions of flapjacks (Livesey et al. 1995a) and bread, both of which were made with barley. Barley was Hordeum cv. maggi from Reedes (Norwich, UK) and had been kibbled and air classified to remove the husk. The kibblings were rolled (1 mm clearance) while dry by Rank Hovis McDougall (Welwyn Garden City, UK) to produce flakes ~2 mm in diameter, which were stored at 
40°C until used. The flakes were incorporated into flapjacks as described previously (Livesey et al. 1995a) or bread. Bread was barley flake (480 g), strong white bread flour (320 g, Sainsbury's, local supermarket), salt (10 g), margarine (32 g), a dried yeast preparation (7 g, McDougall's Fast Action, from a local supermarket) and water (475 mL). Dough was cooked at 215°C in a fan-assisted oven for 25 min. Bread and flapjacks were stored at 20°C until used.
Diet formulations.
Four diets of varied composition (see below) were used in 3-d rotating menus, with individual food items as given in the Appendix. Diets met criteria of nutritional adequacy for vitamins and minerals (Department of Health 1991), and their formulation was facilitated by the use of the Institute of Food Research Computerized Nutritional Information Service (Walker 1992). This service used food composition data from several sources (Englyst et al. 1988 and 1989, Holland et al. 1988 and 1991, Paul and Southgate 1978) and information on reasonable portion sizes (Crawley 1988).
One diet was a weight-reducing, high NSP and low fat diet calculated (Paul and Southgate 1978) to provide ~6 MJ metabolizable energy per subject daily (Table 1). Three other diets were formulated to meet maintenance energy requirements, defined as 1.45 times basal metabolic rate, estimated using Schofield (1985) equations. The first diet had a moderate NSP content (2.1 kJ/100 kJ of gross energy), the second had a high NSP content (4.6 kJ/100 kJ of gross energy) and the third, which was formulated with a high barley content, had an intermediate NSP content (3.5 kJ/100 kJ of gross energy). The first two maintenance diets were formulated to have the same amount of protein, fat and available carbohydrate. Barley in the third maintenance diet was either as bread or as flapjack. There was no difference between the bread and flapjack results; thus the two results were pooled. Energy balance was achieved on the first two maintenance diets by varying the quantity of the total diet consumed. Barley diets were formulated to provide ~8 MJ metabolizable energy and increments of selected non-barley foods (see Appendix) were provided to meet energy requirements. Each diet was eaten by six subjects from a pool of 18.
Subjects.
The study involved 18 healthy male volunteers, aged 21-59 y, weighing 66-100 kg, height 1.72-1.92 m and body mass indices 21-31 kg/m2. Those consuming the weight-reducing diet were slightly overweight initially, having a body mass index between 25 and 31 kg/m2 (Table 2). Subjects were nonsmokers and nonmedicated.
Protocol.
Diets were consumed by volunteers while housed in metabolic suites with individual bedrooms. Those eating the moderate-NSP maintenance diet also ate the high NSP maintenance diet in a randomized crossover design. All diets were consumed for 23-30 d. Duplicate diets were collected and stored frozen at 20°C. Uneaten food remnants were also collected, weighed and stored frozen. Duplicate diets and remnants were each homogenized in water, freeze-dried and stored dry.
Radio-opaque markers were consumed at a rate of 30/d to measure the fractional recovery of feces (Brown 1993). Urine and feces were collected daily. The moderate and high NSP maintenance diets were eaten for 23 d, and excreta were collected for analysis from d 15 to 20 (6 d). Similarly, the high barley and the weight-reducing diets were consumed for 30 d and excreta were collected for analysis during d 15 to 24 (10 d); the 10 d were split into two 5-d periods. Fecal collections continued in all cases for up to another 5 d to allow fecal collections to be associated with the food ingested. Balance periods were distinguished by changes in the shape and size of markers (Cummings et al. 1976).
Fecal samples were frozen each day at 20°C and the number of radio-opaque marker pellets in each sample was counted from X-radiographs. After freeze drying, the feces were powdered and pooled over the collection period and then stored dry. A change in the size and shape of the radio-opaque markers was made at the beginning and end of each collection period.
Twenty-four hour urines were collected onto boric acid (5 g) daily and stored at 20°C. After thawing, representative aliquots from each day of the collection period were pooled and freezed dried.
Analysis.
Powdered duplicate diets, diet remnants and feces were analyzed for dry matter, gross energy, nitrogen, fat, NSP, RS, ash and carbohydrate by difference. Powdered urine was analyzed for gross energy and nitrogen.
Heats of combustion of dried powders were determined in an adiabatic bomb calorimeter (Gallenkamp, Loughborogh, UK) by using a benzoic acid thermochemical standard (Brown 1993). Nitrogen was Kjeldhal nitrogen. Fat was assayed gravimetrically after extraction by refluxing with dichloromethane:methanol (AnalaR) (9:1, v/v) on a Soxtec System HT 1043 extraction unit (Tectator, Hogens, Sweden). Ash was assayed gravimetrically after heating to 480°C for 18 h in a muffle furnace.
RS in diets was the starch remaining after removal of free sugars, dextrins and starch that is available to digestion with 
-amylase (Pancrax, Pains and Byrne, Greenford UK) and pullulanase (to hydrolyze the limit-dextrins) overnight at 42°C and pH 5.2 (Englyst & Cummings, 1988); the RS was solubilized in 7 mol/L KOH and hydrolyzed with amyloglucosidase (Englyst & Kingman, 1991). Glucose was analyzed by the glucose oxidase-peroxidase method (Boehringer Mannheim, Mannheim, Germany) and RS was 0.9 times the weight of glucose. NSP was determined by the gas-chromatography method (Englyst & Cummings, 1988). Total carbohydrate in the diets was calculated by difference, dry matter less ash, protein (N × 6.25) and fat. Available carbohydrate in diets was calculated as total carbohydrate less the NSP and RS.
Calculations.
Metabolizable energy content of diets was calculated in several ways as follows: first, as metabolizable energy [MED; (Eq.1)] determined from the difference in gross energy intake (GE, kJ) and excretion in feces (FE, kJ) and urine (UE, kJ), with corrections to zero-nitrogen balance in the case of the submaintenance diet, at a rate of 30 kJ/g urinary N (Levy et al. 1958, Southgate 1975).
|
(1)
|
Second, metabolizable energy was calculated according to what has been variously called Atwater specific, Handbook 8 or Merrill and Watt (1973) factors [MEM&W (Eq. 2)]. For this procedure, metabolizable energy value was calculated from the protein (P), fat (F), and total carbohydrate (TC) contents of foods and the respective food-specific energy conversion factors (Handbook 8). In practice the composition of individual food items was not determined directly and thus was taken from the food tables; to minimize error we therefore multiplied the result by a ratio: 
(16.7P + 37.7F + 16.7TC) foods/n-1. (16.7P + 37.7F + 16.7TC)diet, where dietary values were determined. This ratio is zero on theoretical grounds but otherwise gives the correct value relative to the Atwater general factor system described below.
|
(2)
|
Third (Eq. 3), using Atwater general factors for protein (P = 6.25 × nitrogen, g), fat (F, g) and total carbohydrate by difference, (TC, g)
|
(3)
|
Fourth (Eq. 4), using European general factors (European Council Directive 1990) for protein, fat and available carbohydrate (AC, g) wherein AC was the difference between TC and the sum of nonstarch polysaccharide (NSP, g) and resistant starch (RS, g).
|
(4)
|
Fifth (Eq. 5), using the modified factorial system suggested by Livesey (1990), wherein the unavailable carbohydrate (UC, g) was the sum of NSP and RS
|
(5)
|
Sixth (Eq. 6), using the general formula proposed by the FDA (1993), which is based on gross energy (GE, kJ) and nitrogen (N, g) intake
|
(6)
|
And seventh (Eq. 7), using the general formula of Livesey (1991a).
|
(7)
|
Fractional bias in each method of calculation was estimated as the ratio of predicted to determined values as given in (Eq. 8).
|
(8)
|
Estimates of the limits of agreement between the various energy assessment systems and the determined value (LAdiet) were obtained as the sum of the departure of absolute bias [from (Eq. 8)] from unity and twice the standard deviation about the bias for diets (Eq. 9) (Altman 1991).
|
(9)
|
Estimates of the imprecision of predictions of energy values for individuals (LAindividuals) were similarly obtained as the sum of the departure of absolute bias from unity and the square root sum of square standard deviation about the bias for diets and square standard deviation about diets for individuals (Eq. 10).
![LA<SUB>individual</SUB>= Absolute (bias −1) + 2 × √ [(<SC>SD</SC><SUB>among diets</SUB>)<SUP>2</SUP>+ (<SC>SD</SC><SUB>among individuals</SUB>)<SUP>2</SUP>]](/content/vol128/issue6/fulltext/986/img010.gif)
|
(10)
|
Statistics.
Results are presented as means and standard deviations or standard error of the means, as specified (±). Values of bias that were significantly different from unity were tested by Student's t test.
| |
RESULTS |
Analyzed composition of diets.
All maintenance diets were essentially similar in protein (range, 15.5-18.7% GE), fat (range, 32-34% GE), available carbohydrate (range, 42-47% GE) and resistant starch (range, 1.7-2.3% GE) but differed ~twofold in non-starch polysaccharide content (range, 2.1- 4.6% GE) (Table 2). Similar comparisons in terms of mass (g/kg) show approximately similar equalities with differences in NSP content (Table 2).
The weight-reducing diet differed from the maintenance diets in the following three respects: the percentage of protein (as %GE) was higher (by 1.1 to 1.3-fold); the percentage of fat was lower (× 0.7-0.75); and the percentage of NSP was much higher (× 1.5-3.3). By contrast, the percentage energy from RS in the weight-reducing diet was similar to that in the maintenance diets (all were in the range from 1.7 to 2.3% GE), as was the percentage of available carbohydrate (all were in the range from 42 to 49% GE).
Body weights.
Predictably, body weights were unchanged by diets consumed at maintenance intakes; at the end of the balance period all values were within 1 ± 2 (group mean dietary value ± SD, n = 6) of those shown in Table 1. Also predictably, significant (P < 0.05) body weight loss occurred in subjects consuming the submaintenance diet (Fig. 1). The mean weight loss was 8.3 (± 1.9, range 6-11.3) kg over the 28 d of treatment. This loss was associated with nitrogen balances of 11 ± 7 and 14 ± 10 g in the two 5-d- balance periods, respectively, and 25 ± 14 g over the whole 10-d balance period. Metabolizable energy in the weight-reducing diet was corrected to zero nitrogen balance by using these imbalances.
View larger version (27K):
[in this window]
[in a new window]
| Fig 1.
Weight loss in volunteers consuming the high non-starch polysaccharide (NSP) weight-reducing diet. Early morning body weights are shown; each datum is for each of the six male volunteers each day consuming the high NSP weight-reducing diet for 28 d as follows: 14 d of adaptation, two consecutive 5-d balance periods (BAL1 and BAL 2, respectively) and a recovery period (REC) during which excreta continued to be collected to recover markers. Food intake was the same in all periods, i.e., restricted.
|
|
Intakes and excretion of energy and substrate intakes.
The studies involving two balance periods yielded similar results in each period, indicating good replication (Table 3). The similar gross energy intakes during each period occurred because volunteers were provided with the same amounts of food in each period (zero degrees of freedom) and there was either nil or negligible food refusal. Volunteers on the weight-reducing diet consumed, on average, 5500 to 6800 kJ less gross energy each day than those consuming the maintenance diets. However, this lower intake was not accompanied by a proportional decrease in fecal and urinary energy excretion; rather fecal energy excretion was larger than we expected for the amount of food eaten.
View this table:
[in this window]
[in a new window]
|
Table 3.
Gross energy, protein (N × 6.25), fat, available carbohydrate (AC), resistant starch (RS) and non-starch polysaccharide (NSP) intake and fecal and urinary energy excretion in men1,2
|
|
Prediction of the metabolizable energy of maintenance diets.
The Atwater specific factor system accurately predicted the energy value of the moderate NSP maintenance diet, but significantly underestimated the energy values of the higher NSP and high barley NSP diets (Table 4). By contrast, the Atwater general factor system always overestimated the determined energy availability of diets consumed at maintenance, although only one case was significant. The European general factor system always underestimated the determined energy availability of these diets. By modifying the general factor systems to assume that the availability of energy from the UC was 8.4 kJ/g, the bias was reduced to <0.5% for all three diets fed at maintenance. The FDA general formula significantly overestimated metabolizable energy of the maintenance diets by up to 8%. By contrast, the general formula (Eq. 7) predicted metabolizable energy of all three maintenance diets to within 2% of the determined value, and within the 95% confidence range of ± 2% published previously (Livesey 1991a).
Prediction of metabolizable energy of the high NSP submaintenance diet.
All of the general factor and formula systems significantly overestimated the metabolizable energy content of the high NSP submaintenance diet by between 3 and 15% of the determined values (Table 4). There was no significant difference, however, between the determined metabolizable energy and the value calculated using Atwater specific factors.
Imprecision of prediction of food energy assessment systems.
The bias over all three maintenance diets and imprecisions of the energy assessment methods for diets and for individuals, based on data obtained at present, are given in Table 5. Among the systems examined, only the modified factorial system and the Livesey general formulae had mean bias <2% of the determined metabolizable energy and limits of agreement within 3%. By definition, all assessment systems were less precise for individuals than for diets [cf. (Eq. 9) vs. (Eq. 10)]. In our volunteers, the limits of agreement were between 2 and 5% greater for the individual than for the diets.
View this table:
[in this window]
[in a new window]
|
Table 5.
Overall bias and precision of prediction of metabolizable energy in 18 men eating
of the three maintenance diets noted in Table 4
|
|
| |
DISCUSSION |
To evaluate the adequacy of food energy assessment systems, a judgement has to be made about what constitutes an adequate prediction of energy availability. For the purpose of the present study, this was deemed to be when predicted metabolizable energy was without progressive bias with increasing dietary fiber intake, when residual bias was within 2% of the experimentally determined value in a group of adults, and when "limits of agreement" (Altman 1991) between determined and calculated values, which we report for the first time, are small. The 2% value for residual bias was chosen because it corresponds to the approximate accuracy that the specific factor system was judged to have reached when applied to mixtures of usually simple foods (basal diet) and a test (often high fiber) food (Merrill and Watt 1973). An accuracy of 2% is seemingly achievable for an energy evaluation system applied to mixed diets when using data from many different laboratories that also used a range of different methods to analyze dietary fiber (Baer et al. 1997, Livesey 1991a). Greater precision is not likely because the interlaboratory variation in heat of combustion (gross energy) determinations in nutritional laboratories (Livesey et al. 1995b) is not much better than 2%, and better than 2% is not expected for theoretical reasons (Horowitz et al. 1980). Moreover, accuracy better than 2% would be meaningless because the error in energy expenditure estimates by indirect calorimetry (Poppitt et al. 1996) is at or about the 2% level. Two percent of metabolizable energy corresponds to an amount of energy that is typically thought to be available from the low-to-moderate amounts of dietary fiber contained in Western mixed diets (British Nutrition Foundation 1990, Livesey 1990) and an energy imbalance that would approximately account for differences in body weight often found when consuming NSP supplements (British Nutrition Foundation 1990, Livesey 1991b). In terms of energy requirement, 2% of metabolizable energy is normally an amount of energy suggested to support the maintenance of each 2 kg of body weight in people of normal weight (Jéquier 1987). A 2% decrease in energy intake may lower the incidence of diabetes, CHD deaths and overall mortality in the morbidly obese by 3-9% (Livesey 1995a). There is therefore good reason to want to aim for an energy evaluation system that achieves an accuracy of 2%. Although this seems very demanding, it should be remembered that it corresponds to between 10 and 30% of fecal energy excretion (Livesey 1990 and 1991a). The 2% limit of bias and the limits of agreement provide a basis against which each energy assessment system can be considered in turn, first for maintenance diets, then for the submaintenance diet.
In the U.S., Handbook 8 is the usual source of energy values used in dietetic practice. The handbook uses specific factors for the protein, fat and total carbohydrate content of specified categories of foods (Merrill and Watt 1973), and such factors are variously called Handbook 8, Merrill and Watt, and Atwater specific factors. The accuracy of the specific factor system when tested by using different categories of foods (often higher fiber) was originally found to be within 2% of the determined values (Merrill and Watt 1973) as follows: fruit and nuts 1 to +2%, legumes 1 to +3%, cereals and dairy products 1 to 2%, vegetables 0 to 1%, mixed diets 0 to 2%, but a high banana diet 5%. Such accuracy of the specific factor system seems difficult to maintain among various laboratories when diets comprise foods high in dietary fiber from a mixture of food categories. Thus, although we found the specific factor approach was without bias with our low fiber diet (Table 4), the approach underestimated our high barley NSP and higher NSP maintenance diets by 5.3 and 3.7%, respectively (Table 4). These differences must be seen in the context of previous comparisons between determined metabolizable energy and the specific factor predictions with higher fiber diets. Five other comparisons have been published since the original validation of the specific factor system and they indicate the following discrepancies: +6%, 3.6%, 2.1%, 0.4% and +5.9% for daily intakes of 93, 86, 65, 56 and 37 g dietary fiber, respectively (Calloway et al. 1978, Göranzon et al. 1983, Miles et al. 1988). If all of these results are given equal weighting, the overall mean bias is 0.7 ± 4.9 ( mean ± SD, n = 7) and the limit of agreement for diets is 10.5%. The mean value (0.7% of the determined metabolizable energy) suggests that the specific factor system is without bias overall, but the large limit of agreement indicates the method to be relatively unreliable. Indeed, values for particular diets frequently (6 out of 7 times) fell outside the 2% limit adopted as a criterion of adequacy. Our finding that the Atwater specific system appreciably underestimates metabolizable energy in two instances is therefore not surprising. The causes of error are difficult to identify because of the complexity of the specific factor system and possible scope for its subjective application. Several potential sources of error have been discussed previously (Life Sciences Research Offices 1983, Merrill and Watt 1973). However, the specific factor system is often accurate for diets to within 6% of the determined metabolizable energy value for high fiber Westernized diets based both on the present data (Table 5) and the above information on all of the higher fiber diets. This is still three times the potential error originally suggested for the specific factor system (Merrill and Watt 1973) and three times the error of the general formula system (Eq. 7) as validated recently in another laboratory (Baer et al. 1997) and in this paper (Table 4). The limits of agreement of the specific system are even greater, 10% (Table 5) or 10.5% for higher fiber diets alone (as calculated above) compared with 3% for the general formular system (Eq. 7) calculated in this paper (Table 5) and 2% calculated previously as a 95% confidence interval (Livesey 1991a). With both the specific factor system and the general formula system (Eq. 7), ~2-5% additional imprecision can be attached when attempting to predict energy availability from diets for individuals (Table 5).
It has been established (Merrill and Watt 1973) that the Atwater general factors predict higher metabolizable energies than the specific factor system, and the present observations are in agreement (Table 4). It has also been well established (Baer et al. 1997, Göranzon et al. 1983, Livesey 1990, Merrill and Watt 1973, Rosado et al. 1992 ) that the Atwater general factor system overestimates the energy values of high fiber diets; again the present results are in agreement (Table 4). In Europe, a departure from the Atwater general factor system arose in which dietary fiber is deducted from the total carbohydrate (European Council Directive 1990); as anticipated from a previous analysis (Livesey 1990), the method results in an underestimation of the determined metabolizable energy (Table 4). The FDA (1993) proposed a general equation for the estimation of metabolizable energy values in U.S. food labeling. The system has not been tested previously, and the present data show this formula (Eq. 5) to well overestimate the metabolizable energy values of all diets (Table 4). All of these systems showed either positive or negative bias with increasing dietary fiber intake.
A modified factorial system (Eq. 5) has been proposed (Livesey 1990) in which the metabolizable energy value of dietary fiber is taken to be 8.4 kJ/g; this suggestion arose from an analysis of energy availability from 17 diets of varied dietary fiber content but has not previously been tested with independently obtained energy balance data. The present results (Table 4) appear to validate the modified system, showing it to cope with variation in the dietary fiber content of the diet. However, it may not apply to very low fat diets. Since performing the present studies, Baer et al. (1997) have published information on nine diets at three levels of dietary fiber intake and three levels of fat intake; our application of the modified general factors (Eq. 5) to their data indicated, again, a lack of bias with increase in fiber intake, but we calculated a positive bias (between 3 and 7%) for their very low fat diets. On the same nine diets, the Livesey (1991a) general formula (Eq. 7) was accurate to within 2% (Baer et al. 1997). This general formula (Eq. 6) was developed on 43 diets of varied dietary fiber content, and the present observations (Table 4) confirm its validity to within 2% for diets of variable fiber content as assessed from the sum of NSP and RS. The formula was developed originally with the use of data for which dietary fiber had been measured variously as Southgate, Prosky, Asp, Theander, Meuser and neutral detergent fiber (see Livesey 1990 and 1991a). Although the formula seems relatively robust against differences in methods for dietary fiber analysis, we do not recommend the neutral detergent fiber method (mostly insoluble fiber) or the NSP method alone (because it escapes a measure of resistant starch).
We included one maintenance diet with a high proportion of cooked dry rolled barley. Barley kibbling has been reported by Miller and Judd (1983) to cause losses of energy to feces that Livesey (1990 and 1991) find to be larger than expected on the basis of its dietary fiber content. However, no such effect was observed in this study. We suggest that this "kibbling effect" is due to the presence of intact tissue from whole-grain cereal. Elevated losses from the small to the large intestine result from consuming barley particles sufficiently large to elevate the ileal excretion of both lipid and starch encapsulated in barley plant cell walls (Livesey et al. 1995a). Nevertheless, wet rolled barley (1 mm × 4 mm) does not elevate ileal energy losses in amounts comparable to those fecal losses observed by Miller and Judd (1983), and the present dry-rolled barley of even smaller particle size (1 mm × 2 mm) was without any notable effect when including RS in the measure of unavailable carbohydrate. We suspect also that dry rolling of barley makes the starch more susceptible to digestion than when the barley is rolled wet. The result is consistent with the observation that small changes in particle size of rye are without influence on energy availability (Wisker et al. 1996a and 1996b).
The metabolizable energy of the high NSP submaintenance diet was very low compared with predictions by all of the general factor and formula systems tested (Table 4). This was not due to a lack of correction of the determined metabolizable energy to zero nitrogen balance, because the comparisons of determined and predicted values were made after correction to zero nitrogen balance. The observation emphasizes that the general formula (Livesey 1991) applies to maintenance diets only. With submaintenance diets, it seems that a correction would be required to predict energy availability accurately. A first estimate of such a correction factor can be made on the basis of the present observations, which suggest that metabolizable energy should be decreased by 1 ± 0.3% (mean ± SEM) per 10% fall below maintenance energy requirements. By contrast, the availability of metabolizable energy from the high NSP submaintenance diet was well predicted by the specific factor system (Table 4). This was surprising given that the system was developed for maintenance diets; the close correspondence might be interpreted in two ways. First, the specific system potentially is more widely applicable than previously thought. Second, the agreement between the determined and the specific factor prediction is simply fortuitous. There is no basis at present to distinguish between the two possibilities; given a limit of agreement with determined values of only 10% (Table 5) for maintenance diets, the second interpretation is not unlikely.
In conclusion, information available at present suggests that the most reliable system for the prediction of metabolizable energy availability from maintenance diets is that of Livesey (1991a) (Eq. 7). General factor systems may be made to cope with variation in dietary fiber intake, but on the basis of data published by Baer et al (1997), they may remain suspect in considering extremes of other dietary compositions. Other criticisms have recently been advanced against the use of general factor methods (Livesey 1995b). For mixed diets, the general formula (Livesey 1991) is at least as accurate as was originally suggested for the Atwater specific factor system, and according to more recent data, could result in three times less error. There is at least a case for concern over the complexity and reliability of the Atwater specific factor system as a "gold standard." For food labeling purposes, the general formula approach appears highly reliable (Baer et al. 1997, Livesey 1991a, Table 4), probably because of the advantages that have been outlined previously (Livesey 1995b). For our weight-reducing, low fat, high fiber and low energy diet, less metabolizable energy is available than predicted by this formula, which emphasizes the fact that, at present, food energy assessment systems reliably apply to maintenance diets only, unless adjustments are made. Finally, none of the energy assessment systems account for fermentation losses (hydrogen, methane, heat of fermentation, inefficient yield of ATP for short-chain fatty acid products of fermentation and for amino acids from protein), for which further adjustments would have to be made (Brown et al. 1993, Livesey 1995b, Poppitt et al. 1996).
| |
ACKNOWLEDGMENTS |
We wish to express our thanks to both the Biotechnology and Biological Sciences Research Council (BBSRC) and the Medical Research Council (MRC) for providing facilities.
![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
,
Abstract
Introduction
Abstract
comparison of the experimentally determined and the calculated metabolizable energy of sixteen diets.
Zeitschrift Fur Lebensmittel-Untersuchung Und-Forschung
1993;
197:233-238
