Contents - Previous - Next

3.5. Starvation in man and other species

About 80-100 years ago various workers, notably VOIT (1901), observed differences in P% in various species undergoing starvation. The general concept that obesity is associated with a lower value of P% crosses the species barrier, and evidence for this can even be found during early starvation (Table 4). The pig, which has abundant fat, has a lower value for P% than typical humans and rabbits that are leaner. Similarly, in birds with abundant fat the P% is lower than in lean hens, especially those with virtually no fat (Table 4). Therefore, the concept that protein oxidation contributes to a variable proportion of energy expenditure, depending on the degree of adiposity, originated almost a century ago. The analysis in this chapter provides the most detailed evidence so far that the same concept applies to the human species. One can also predict that P% during prolonged starvation is related to the percent body fat, and evidence for the relationship is present in Figure 11.

The above relationships raise important questions about the metabolic processes that regulate the proportion of energy derived from protein oxidation. What are the signals that are responsible for preserving protein more effectively in obese individuals? This still remains a major problem in metabolic regulation. Although ketone bodies and fatty acids may spare protein (SHERWIN et al., 1975; NAIR et al., 1988; PAWAN and SEMPLE, 1983; TESSARI et al., 1986; LOWELL and GOODMAN, 1987), there is insufficient information about the kinetic changes or even about the circulating concentration of these metabolites and hormones during prolonged starvation, to enable comparisons to be made between lean and obese individuals. Most of the studies of prolonged starvation in lean humans were undertaken before 1925, when methods for measuring circulating hormones and metabolites such as non-essential fatty acids and ketone bodies were either inadequate or non-existent. In short-term starvation, differences in the circulating concentration of ketone bodies exist between men and women, but the differences between lean and obese subjects are particularly pronounced (ELIA, 1991). Paradoxically, the values during early starvation are greater in lean individuals than obese subjects. Whether the reverse situation occurs during prolonged starvation is uncertain.

4. Hypocaloric dieting

4.1. Duration of dieting
4.2. Protein and energy intake
4.3. Body composition
4.4. Exercise

The prevalence of obesity (BMI >25) in Western societies may be more than 30%, but most of the affected individuals are mildly to moderately obese, and only a small proportion is grossly obese. Not surprisingly, most individuals attempting to lose weight are mildly to moderately overweight, and some have normal or near normal BMI. The question therefore arises as to whether there are intrinsic differences in the response to various hypocaloric diets of individuals of different body composition, and if so, what are the optimal diets that should be prescribed. The field has been controversial and dietary compliance, acceptability, and well-being remain important issues. In the brief analysis that follows, an attempt will be made to review critically the importance of body composition in the context of some of the other factors that affect protein-energy interrelationships during hypocaloric dieting.

Table 4. N loss during starvation and the contribution of protein oxidation to energy expenditure

Animal (weight in kg)

Length of starvation (days)

Energy from protein (% BMR)


Pig (115 kg)



Man (58 kg)



Dog (8.7/25.6 kg)



Rabbit (2.4/2.5 kg)




Geese (3.6/3.7 kg)



Hen (9% fat) (1.0 kg)



Hen (210 fat) (1.7 kg)



Data from VOIT, 1901; RUBNER, 1883; MEISSL, 1886; and BENEDICT, 1915.

A negative N balance during hypocaloric dieting is not necessarily detrimental. This is because obese individuals have more fat-free mass than lean individuals. Although there is substantial interindividual variability in the composition of excess weight, it is generally considered to consist of about three parts fat-free tissue to one part fat, over a wide range of body weight (WEBSTER et al., 1984; GARROW, 1981). FORBES (1987) has suggested a curvilinear relationship between lean body mass and fat mass, but this is largely due to the inclusion of patients with anorexia nervosa in his analysis. Without this patient group, the relationship between fat and fat-free mass is close to linearity.

Studies in subjects ingesting a fixed amount of protein (above minimum required for maintenance), have confirmed that excess energy intake causes a positive N balance, which persists as long as the hypercaloric feeding persists (ELIA, 1982). Grossly obese individuals (FORBES, 1987; JAMES et al., 1978) may have over 30% more fat-free mass than lean individuals of the same height. In the example shown in Figure 3, the obese individual weighting 140 kg has a fat-free mass that is 29% greater than the 70 kg man. Obese individuals appear to have more muscle and bone than lean individuals, and these help support and move the excess body weight. Obese subjects have large vascular volumes and larger hearts, which are necessary to pump more blood around larger bodies, especially during weight-bearing activities. Obese individuals may also have visceromegaly (NAEYE and ROODE, 1970). Finally, adipose tissue contains about 15% water and 5% protein (SNYDER et al., 1975), so that the presence of excess adipose tissue is associated with excess fat-free tissue (although in obese individuals adipose tissue has a smaller fat-free component than in lean individuals).

Since all the above tend to increase fat-free tissue in the obese, dieting is expected to reduce the amount of fat-free tissue and cause a negative N balance. This loss should be regarded as physiological, although excessive loss of lean tissue, especially when it is associated with loss of body function, is clearly not desirable.

Figure 12 shows that there is substantial variability in the composition of weight loss in subjects undergoing dieting for 3-26 weeks. Even greater variability occurs during the first 3 weeks. An understanding of the factors that contribute to this variability could have important implications for dietary recommendations. Therefore an attempt is made below to consider some of the factors individually.

Figure 12. Contribution of fat to weight loss in groups of individuals (BMI 27-50 g/m2) receiving hypocaloric diets (330-1900 kcal/d - only four studies with intakes > 1000 kcal/d and 4.8-18.4 g N/d) for periods ranging from 3-26 weeks.

Based on data of DAVIES et al., 1989; BARROWS and SNOOK, 1987; FRICKER et al., 1991; FOSTER et al., 1990; ELLIOT et al., 1989; RAVUSSIN et al., 1985; WEBB et al., 1983; BJÖRNTORP et al., 1975; HENDLER and BONDE, 1988; BOGARDUS et al., 1981; YOUNG et al., 1971; PHINNEY et al., 1980; LANTIGUA et al., 1980; BELCO et al., 1985; YANG and VAN ITALLIE, 1984; VASWANI et al., 1983; Woo et al., 1982; and COXON et al., 1991. Various body composition techniques were employed: densitometry, total body water, total body potassium, total body nitrogen. Nitrogen balance was included in some of the techniques and such studies are indicated by open bars.

Figure 13. Daily N balances in groups of obese individuals at various times after the start of hypocaloric dieting (320-720 kcal/d, 5-21 g N/d).

Subjects within each group received the same hypocaloric diet. Based on HENDLER and BONDE, 1988 (); WILSON and LAMBERTS, 1979 (O); HOFFER et al., 1984 (l); and YANG et al., 1981 (n).

4.1. Duration of dieting

Figure 13 shows the daily N balance in groups of subjects receiving low-calorie diets at various times after the start of dieting. Although there is substantial variability in results among various groups of subjects, there is a constant and progressive reduction in N excretion during the first 4 weeks. Therefore, although each group of subjects is receiving the same diet, different conclusions about the contribution of lean tissues to loss of body weight can be drawn if different time points are chosen. Indeed, the N loss during the first week is frequently two-fold greater than the one during the third or fourth week. Calculations based on the results of several hypocaloric studies suggest that there is a progressive decrease in the loss of N/kg of body weight (Figure 14). This implies that the contribution of lean tissues to the weight loss becomes smaller with time.

If fat-free tissue, which contains about 32 g N/kg, were to account for 25% of the weight loss, as suggested by cross-sectional data of body composition in non-fasting lean and obese subjects (see above), then 8 g N are expected to be lost for every kg loss of body weight (assuming that the lean tissue constituents are lost in approximately similar proportions as they exist in the body). Figure 14 suggests that values close to the 8 g N/kg weight loss are frequently achieved after the first 2-3 weeks of dieting.

In summary, conclusions about protein-energy interrelationships obtained during the early period of dieting may not apply to the subsequent weeks.

Figure 14. Cumulative N loss/cumulative weight loss during hypocaloric dieting.

Based on WILSON and LAMBERTS, 1979 (320 kcal/d,5 g N/d) (•); HOFFER et al., 1984 (500 kcal/d, 7 g N/d) (n); and YANG et al., 1981 (720 kcal/d, 21 g N/d) (). The results from this study with 630 kcal/d and 10.5 g N/d are similar but are not shown.

Figure 15. N balance at 3-5 weeks in obese individuals according to N intake and energy intake.

Based on CONTALDO et al., 1980; HENDLER and BONDE, 1988; PASSMORE et al., 1958 (only subjects on stable diets); LANTIGUA et al., 1980; DURANT et al., 1980; YOUNG et al., 1971; and ELLA (unpublished). The curves are drawn visually.

4.2. Protein and energy intake

Figure 14 shows that the cumulative N loss in groups of obese individuals with a body mass index greater than 35 receiving very low calorie diets, containing 300-400 kcal/d, and a variable N intake. As the N intake increases, N loss becomes smaller, but there is little or no effect between 6.5 and 17.0 g N/d. Although the higher N losses in Figure 14 are attributed to low protein intakes, which fall below the minimum safe requirement for normal individuals (FAO/WHO/UNU, 1985), it is possible that in some studies lack of other dietary constituents such as potassium and phosphate may have contributed to the loss (RUDMAN et al., 1975). Figure 15 suggests that when the N intake in obese individuals is kept constant, N balance is improved by increasing the energy intake. When the energy intake is kept constant, N balance is improved by increasing the N intake although there appears to be little improvement above an N intake of 6-7 g N/d. This is analogous to the situation in normal man (CALLOWAY and SPECTOR, 1954; ELIA, 1982).

This conclusion is supported by individual studies in obese subjects. HENDLER and BONDE (1988) reported no improvement in N balance when the N intake increased from 6.5 to 16.5 g N/d while the energy intake was kept constant at 440 kcal/d. Other studies (YANG et al., 1981; YANG and VAN ITALLIE, 1984) also suggest that there is no improvement in N balance when the N intake is increased from about 10 to 20 g N/d while the energy intake was maintained at 620-720 kcal/d, or when it is increased from about 8 to 16 g N/d while the energy intake was maintained at 400 kcal/d (DE HAVEN et al., 1980, although in this study a transient difference was observed during the first week; lack of K may have been a feature of this study). However, the study of HOFFER et al., (1984) suggested a significant improvement in N balance when the N intake was increased from 7.0 to 13.6 g N/d while energy intake was maintained at 500-550 kcal/d.

Figure 16 shows the contribution of protein oxidation to energy expenditure (% BMR) in groups of obese individuals with a BMI of more than 35. As the protein intake increases, the proportion of energy derived from protein oxidation also increases. However, the endogenous protein oxidation (negative N balance) only makes a substantial contribution to total protein oxidation at the lower levels of N intake (Figure 16).

4.3. Body composition

Since the composition of weight loss in lean individuals undergoing total starvation differs from that in obese individuals, the question arises as to whether the same applies during ingestion of low-calorie diets. FORBES (1987) has indeed suggested that during hypocaloric dieting the contribution of the fat-free body to loss of body weight is greater in lean subjects than obese subjects.

Figure 16. Contribution of protein oxidation to basal metabolic rate (BMR) during the second week of dieting in obese subjects receiving different protein (N) intakes but fixed energy intake (300-440 kcal/d).

The bars represent the uncertainty of BMR (80-120% of predicted BMR). The data represent group measurement made during the second week of dieting. Both curves are displaced upwards during the first week of dieting (data not shown). The endogenous oxidation represents oxidation of endogenous protein in subjects in negative N balance.

The calculations of protein oxidation are based only on urine N measurements since the N lost in faeces and skin is largely protein which is unoxidised. Calculated using data of MARLISS et al., l978; DAVIES et al., l989; HENDLER and BONDE, 1988; PASSMORE et al., 1958; ROZEN et al., 1986; and WILSON et al., 1979.

Figure 17. Fraction of weight loss due to lean body mass (D LBM:D W ratio) in underfeeding experiments lasting for at least four weeks.

Data points are averages ( SEM) of three or more subjects (various body composition techniques). Reproduced from FORBES, 1987.

Figure 17 produced by FORBES (1987) illustrates this for subjects receiving different energy intakes. The more obese the subject, the smaller the contribution of fat-free mass to the loss of body weight. Since this has major potential implications for hypocaloric dieting in individuals with different initial body composition, it is important to consider briefly the origin and some of the limitations of the results shown in Figure 17.

1. Age. Some of the studies represented in Figure 17 (FORBES, 1987) are studies of adolescents while others are of adults.

2. Protein intake. Figure 17 (FORBES, 1987) specifies the energy intake of the subjects but not the protein intake. Indeed, some of the studies used by Forbes (1987) to construct Figure 17, employed a N intake as high as 18.4 g N/d (YOUNG et al., 1971), while in others it was less than half this (KEYS et al., 1950). In yet other studies the N intake was not specified (BELKO et al., 1985). The source of the N was also variable, ranging from collagen hydrolysates supplemented with tryptophan (LANTIGUA et al., 1980) to protein from normal food (PHINNEY et al., 1980).

3. Variable intake of various dietary constituents. For example, supplements of minerals and micronutrients were used in some studies but not others. There was also a variable ratio of carbohydrate to fat intake. There has been a long-standing controversy about whether high- or low-carbohydrate diets are superior in sparing protein in both normal subjects and obese individuals undergoing hypocaloric dieting. Some workers have favoured the ketogenic (low-carbohydrate) diets on the basis that ketone bodies spare protein during caloric deprivation. Other workers have found little or no difference between ketogenic and non-ketogenic diets. Yet, other workers favour the non-ketogenic diets since they have observed that isocaloric replacement of fat for carbohydrate is associated with improved N balance (see VASQUEZ and ADIBI, 1992 for discussion).

4. Length of study. The studies included in Figure 17 were based on measurements made in subjects who had dieted for periods ranging from 4 weeks to over 6 months. It is possible that some compensatory changes occur with time. For example, in the study of BELKO et al. (1985), in which the subjects ingested 1250 kcal/d, the contribution of fat-free mass to weight loss (combined results based on body composition measurements by densitometry) was 54% at 3 weeks, 40% at 6 weeks and 26% at 9 weeks. A change in the form of Figure 17 would result if the earlier values were used instead of the 9-week value.

5. Exercise. This is another variable which is considered separately below.

There are also some studies with hypocaloric diets that, at first sight, might appear to contradict the concept that obesity is associated with better protein economy than leanness during hypocaloric dieting.

The N balances during the first 2 weeks of dieting with 550-700 kcal/d and 10-13.5 g N/d in young adult subjects with different body composition have been reported as follows:

- 1.1 g N/d in subjects (n=49) with a BMI of < 25 (106.5+-1.7% of Metropolitan Life Insurance Company tables) (SEROG et al., 1982);

- 1.5 in a group of subjects (n = 9F) with a BMI of 25-35 (HOFFER et al., 1984); and

- 6.4 (n=3M) and 6.9 (n=3M) in grossly obese subjects with a BMI > 40 (YANG and VAN ITALLIE, 1984; YANG et al., 1981).

In these studies the more obese subjects had the larger rather than the smaller N losses.

Another recent study (COXON et al., 1991; MORGAN et al., 1991) employing a variety of body composition techniques (water dilution, total body potassium, total body N. hydrodensitometry and dual energy X-ray absorptiometry) has shown no relationship between composition of weight loss and initial BMI (22.1-42.8) of subjects receiving a very low calorie diet (405 kcal and 42 g protein per day) for 11 weeks. The composition of the weight loss was reported to be 19% fat-free mass (4.2% protein) in those with an initial BMI of 32.542.8, 23% fat-free mass (5.12% protein) in those with a BMI of 25.032.5, and 19% fat-free mass (4.62% protein) in the individual with a BMI of 22.1. The weight loss was 16.2 2.4 kg and the BMI decreased by 5-6 points, irrespective of initial BMI. For example, the leanest subject decreased her weight from 65.6 to 49.5 kg, and her BMI from 22.2 to 16.9.

The apparent discrepancy between these last two studies on the one hand, and Figure 17 on the other, requires further consideration, because of the enormous implications for the use of low-calorie diet in individuals with a different body composition. Two important aspects are quantitative. The first is the magnitude of the N loss floss of fat-free body). Figure 8 shows that during total starvation in the obese, N excretion can be as low as 4 g N/d, especially as the fast progresses. Figure 13 suggests that the use of hypocaloric diets in obese subjects can reduce the N loss further. Since some N loss is normally expected when there is a large negative energy balance (e.g., loss of three parts fat to one part fat-free mass implies a negative N balance of about 1-2 g N/1000-2000 kcal energy deficiency), the extra protein-sparing effect of very low calorie diets in obese individuals is relatively small, especially during prolonged dieting. Therefore, to demonstrate this effect it is necessary to use accurate and precise N-balance techniques, which assume good dietary compliance and complete urine collections. Although simple in principle, the N-balance technique may be subject to a number of practical difficulties.

The second quantitative aspect concerns the degree of obesity and its relationship to loss of fat-free tissue during hypocaloric dieting. It can be seen that both for total starvation (Figure 11) and hypocaloric dieting (Figure 17) there is a curvilinear relationship between the two variables, which is predictable from models of survival during dietary deprivation (e.g., see Figures 3 and 11 for starvation). In particular, there is a rapid decrease in N loss (or loss of fat-free body) as very lean subjects increase their body fat, but the effect becomes progressively smaller as the degree of obesity increases. In Figure 17, the only convincing data for high loss of fat-free mass in lean individuals receiving hypocaloric diets with an intake of greater than 1000 kcal/d concerns individuals with less than 10 kg of body fat (the upper left point corresponds to only about 5 kg of body fat). One set of data presumably represents the results of KEYS et al. (1950) which were used in the analysis. Here, the initial body fat estimated by hydrodensitometry and water dilution was 9.8 kg (14% body fat) in individuals with a BMI of 21.7. It is unlikely that normal individuals with this or a lower BMI would wish to lose further weight by hypocaloric dieting.

If we now return to the reports of COXON et al. (1991) and MORGAN et al. (1991), in which the leanest woman with a BMI of 22.1 (close to the ideal body weight) dieted for 11 weeks, her initial % body fat is likely to have been almost two-fold greater than that of male subjects of the Minnesota study, who had a similar BMI (KEYS et al., 1950) (at a given BMI women have more % body fat than men). It is also noteworthy that the very acceptable N balances reported by SEROG et al. (1982) (see above) were in individuals of unspecified gender who were 106% of their ideal body weight. With the use of subjects with such widely different body composition it is not too surprising that the conclusions of various studies are different.

There are also a number of assumptions about individual body composition techniques which may be open to criticism, especially during starvation or dieting. The uncertainties include the hydration fraction of fat-free mass (water dilution techniques), density of fat-free mass (hydrodensitometry) and potassium concentration and K/N ratio of fat-free mass (total body potassium). For example, the loss of K during starvation does not parallel N loss. A more detailed discussion of the problems is given elsewhere (ELIA, 1992). The use of a combination of body composition techniques in appropriately structured models of body composition improves confidence in the results without compromising precision (ELIA, 1992; FULLER et al., 1992).

4.4. Exercise

There is increasing evidence that exercise reduces the contribution of fat-free mass to weight loss (HILL et al., 1987; HEYMSFIELD et al., 1989; PAVLOU et al., 1985; WELTMAN et al., 1980; VAN DALE et al., 1989; BALLOR et al., 1988; LENNON et al., 1985; BELKO et al., 1987) although two studies report essentially no effect (WIRTH et al., 1987; HAMMER et al., 1989) and another suggests the opposite (PAVLOU et al., 1989). It is possible that a variable degree of exercise has contributed to some of the variability in the results of some of the hypocaloric feeding studies described above.

A detailed review of the effect of low-calorie diet (oral or intravenous) in clinical practice is beyond the scope of this paper. However, low-calorie protein-sparing regimens administered orally or intravenously have been used with apparently good results in a variety of hospitalized patients with low-grade disease.

There are few data on the influence of initial body composition on the protein-sparing effect of low-calorie diets in patients with severe acute disease. However, a recent study reported that obese subjects who are acutely and severely injured, have greater rather than smaller N losses than lean individuals with the same injury severity (JEEVANANDAM et al., 1991). Given the nature of the multiple injuries sustained by the subjects studied by JEEVANANDAM et al., (1991), it is surprising that the N losses were, in some subjects, as low as 7-11 g/d (no intake of N), although renal impairment may have contributed.

Contents - Previous - Next