Contents - Previous - Next

10. Overfeeding studies

10.1. Early studies
10.2. Prolonged overfeeding

Since the last FAO/WHO/UNU Committee Report (1985), there has been a substantial increase in our understanding of the degree to which adults can respond to overfeeding by increasing their metabolic rate. Nevertheless, there is an increasing tendency to misquote earlier studies, some of which will therefore be considered here. It has been claimed that some adults can vary their food intake over a wide range and yet show only modest changes in weight.

10.1. Early studies

The classic descriptions are those of NEUMANN in 1902 and GULICK in 1922. Neumann monitored his own body weight for a year and varied his calorie intake so that in three major periods during the year he was on 1766,2199 and 2403 kcal per day. On this variable intake his weight remained essentially constant. It should be noted, however, that the capacity to maintain weight stability was displayed with energy intakes which were increased or decreased by only about 15%.

Gulick made a much more detailed study, monitoring not only his food intake but recording very accurately the degree to which his physical activity changed over a period of 370 days. During this time, he varied his intake from 1974 to 4113 kcal/d. He was able to maintain his weight on 2750, 3200 or 3500 kcal/d and not at the extremes of intake. The experiment was also designed so that Gulick could find the minimum amount of food required to maintain an approximately normal body weight. On an intake of 2750 kcal/d, he had an average weight of 62.4 kg over a period of two months, from March to May 1916. He then increased his food intake to 3480 kcal/d and, by the following year, he had gained further on 4113 kcal/d to achieve a weight of 74.3 kg. Although it is claimed that he was unable to show any change in activity or basal metabolic rate, detailed examination of the paper shows that, in practice, a metabolic rate measurement was only undertaken when he was on a high diet. FORBES (1984; 1986) has clearly demonstrated that there is nothing remarkable about the Gulick and Neumann data and has shown the consistency of response to overfeeding in a group of 15 adults.

Figure 2. Mean weekly intakes and expenditures in kilocalories per day (Experiment IV). Reproduced from p. 88 of Stock's thesis where it is claimed that no discernable change in total body potassium, total body water, 24-hour creatinine, subcutaneous fat or pedometer readings of spontaneous physical activity occurred, but no data are presented to substantiate the claim. Weeks 1 and 5 were control periods before and after the 3-week overfeeding study.

MILLER and MUMFORD (1967) overfed 49 subjects with a variety of foods, and Figure 2 reproduces what they claim to be their results. Unfortunately, Miller in his papers never provided formal evidence on any techniques, nor on the degree of individual variability which he claims is present. One therefore has recourse to his Ph.D. students' theses for further details (STOCK, 1970; WISE, 1974). Even here, however, detailed analysis is difficult because no details are given of the amount of food ingested by each student, nor were values for the digestible energy intake given in Stock's thesis which deals with the low-protein and high-protein overfeeding experiments. Figure 2 reproduces the only graph where data on intake and output are both available. Stock claims that the BMR does not change (Table 9), but that there is a marked interaction between postprandial thermogenesis and exercising which explains the ability to buffer excess energy intakes (Figure 3). He notes that a 2-week period of substantial overfeeding is needed to achieve adaptation. GARROW (1978) also concluded that substantial overfeeding is needed before any increase in BMR becomes apparent. Stock's studies were made with direct measurements of oxygen consumption for half the daytime; night-time measurements were also taken so that the data could be calculated for 24-hour energy output. The interaction of meals with exercise (Figure 3) is striking. If true, this would reflect a remarkable capacity to dissipate heat, but there are no data which confirm this, except in the form of a Lancet letter from MILLER and WISE (1975). This other study is the only one from the Miller group where a positive interaction between exercise and diet has been obtained. Other studies listed in Wise's Ph.D. thesis do not confirm the published findings and there remains considerable doubt in the literature about the interaction of diet and physical exercise despite Stock's claim that even mild degrees of exercise enhance the thermogenesis from a meal. Studies by GARBY, LAMMERT and NIELSEN (1984) have specifically tested the Miller and Wise suggestion and failed to substantiate it.

Figure 3. Thermic responses to meals of different sizes (Experiment III). From STOCK, 1970.

Table 9. The effect of prolonged overeating on BMR and the interaction of food and exercise

BMR changes on overfeeding


BMR before

after overfeeding

% change





+ 7.7




- 8.3




+ 5.8




- 7.9




+ 7.4





Fasting kcal/min

Postprandial rise 1 hour after a meal kcal/min







Each value is the mean of 24 determinations on six subjects.
No variance values are given.
From STOCK (1970).

10.2. Prolonged overfeeding

Some short-term overfeeding studies do show a small increase in BMR (e.g., GOLDMAN et al., 1975), but the best of the "old" studies on metabolic adaptation to overfeeding are those conducted by SIMS (1976) on Vermont prisoners and volunteers. In some of these overfeeding studies with mixed meals, a remarkable change in energy homeostasis is claimed. The men chosen for the study were thin individuals without a family history of obesity. Continuous monitoring of food intake and physical activity over a 40-week study-period showed that some men initially ingested approximately 3000 kcal/d, but after overfeeding for months with 6-8000 kcal/d they had gained only 6 kg and now required 5750 kcal to maintain this excess weight. This adaptation occurred despite the men having initially engaged in quite strenuous physical activity on the 3000 kcal diet but then reducing exercise to an absolute minimum in an unsuccessful attempt to become obese. On this basis, the data imply that some feature of metabolism is so flexible that an effective doubling of metabolic activity is possible. Malabsorption was excluded in these long-term studies by Sims but, unfortunately, there was no confirmatory data with measurements of energy expenditure, such as those presented by Stock in his 3-week overfeeding studies (Figure 2). Sims' overfeeding studies therefore need checking by whole-body calorimetry, and his data do not accord with later experience; the prisoners also had a well-recognized tendency to manipulate their investigators when it seemed advantageous (GOLDMAN, personal communication). SIMS' later analysis of the work (1986) is notable for its caution.

Table 10. Effects of overfeeding on energy expenditure measured over periods of 24 hours in a calorimeter chamber or suit


Overfeeding as % maintenance expenditure

Type of food

Duration overfeeding Days

Energy expenditure


Fraction of excess energy %

24-h increase %

4M, 4F






DAUNCEY (1980)





8, 11

5.6, 6.4


4M, 5F

D 1 Mcal or




7.4 2.7

WEBB & ANNIS (1983)






SCHUTZ et al. (1985)

2M, 14F






DE BOER (1985)





7.2 3.0

RAVUSSIN et al. (1985)


24* (15.3)

21.5** 14.0

8F obese




8.2 3.9

3.7 2.1

ZED & JAMES (1986)

8F lean




14.2 3.3

8.1 2.0





19.4 6.8

9.7 2.6

BISDEE et al. (1987)


22.5 8.2

10.9 6.7

Note: Much of these data are recalculated as mean SD from the primary data where given, the excess energy-load being specified in relation to the energy expenditure measured in the calorimeter unless other information is provided. The Webb & Annis data are given for the whole group of normal and overweight volunteers since no group distinction was found, whereas the lean and obese data provided by Zed & James are calculated separately to show the differences in response.

* Value provided by authors after adjustment for the excess physical activity on overfeeding of one subject.

** No adjustment made for one volunteer's excess activity, but the group mean of the other four volunteers is 15.3 1.8.

More recent studies of energy adaptation have concentrated on relatively short-term studies where young or middle-aged men and women have been deliberately overfed. Two of the most detailed so far have been those undertaken by WEBB and ANNIS (1983) and by RAVUSSIN et al. (1985).

The Ravussin study involved establishing five young men first on a weight-maintenance diet before overfeeding them for nine days with 1.6 times their maintenance energy intake. The BMR and 24-hour energy expenditure were measured, with estimates of physical activity and the thermic response to meals. Only about a quarter of the excess energy was dissipated in increased energy expenditure, a value very similar to that observed in other overfeeding studies which involved carbohydrate overfeeding (SCHUTZ, ACHESON and JÉQUIER, 1985). About a third of the increased 24-hour energy expenditure was ascribed to a rise in the BMR, and about 30% of the response reflected an enhanced thermic response to the meal (presumably the 7-10% cost of absorbing and processing the extra food energy) which, the authors note, was not different from the apparent processing costs under normal weight-maintenance conditions. The remainder reflected their estimate of enhanced physical activity and changes in the metabolic cost of physical activity. No separate estimate of the metabolic cost of physical activity was made, but the authors conclude that the data do not provide any evidence for Luxuskonsumption which could be interpreted as a change in metabolic efficiency.

Table 10 shows that the metabolic response to mixed overfeeding may be comparable with that observed with selective carbohydrate overload, but the response to a selective increase in fat energy is less. This reduced thermogenetic response and the suggestion of a genetic nutrient interaction in displaying differences between lean and obese individuals is interesting, but not a key feature relating to the Sukhatme and Margen hypothesis.

Figure 4 Schematic presentation of research protocols to be used by the Consortium for Research on Energy Adaptation in Man (CREAM):

A - Community study on metabolic adaptation in energy

B - Community study on metabolic adaptation in energy

These calorimetric studies, including those listed in Table 10 (DAUNCEY, 1980; DALLOSSO and JAMES, 1984; ZED and JAMES, 1986; DE BOER, 1985), have included modest physical exercise of the type encountered in everyday life, and there has been little evidence of any appreciable change in the metabolic efficiency of physical activity sufficient to make a measurable difference in 24-hour energy expenditure. Recently a number of studies have been undertaken suggesting that intense or prolonged physical activity can lead to some increase in metabolic rate over a sustained period of several hours. This somewhat unusual finding may reflect the stress effects of the intense activity on the sympathetic recovery phase. It is questionable, however, whether this degree of activity is of importance in everyday life.

11. Attempts to test the Sukhatme-Margen hypothesis(es)

Currently a proposal is emerging to form a Consortium for Research on Energy Adaptation in Man (CREAM) involving the Bangalore, Rome, Wageningen and Aberdeen groups in detailed calorimetry, e.g., by protocol VI of the Beaton-Viteri proposals, but tying this in with "field studies" which are cross-referenced to the biological analyses by also being conducted in Bangalore. These field studies would involve the selective overfeeding of both thin and normal-weight individuals in three less developed countries, where food intake is controlled in a hostel environment but where activity is "free". Behavioural and other tests could be added to the protocol which would be based on protocol IV.

In this way it is hoped to cope with the Sukhatme-Margen hypothesis as set out by using protocol VI in developed countries but also address questions (b), (c) and perhaps (e) (listed above) in the field studies using protocol IV.

To simplify reference, Figure 4 schematically shows both protocols IV and VI, with some modifications to take account of the potential for more rigorous control of food intake in protocol IV and the usefulness of having data as nearly comparable with protocol VI as possible. The multiple BMR readings and dietary thermogenesis tests could be used to provide reasonable precision which could well reduce the numbers originally suggested, particularly as food intake would be better controlled. The ability to use the D2O18 technique under tropical circumstances needs to be considered carefully, but there is an opportunity now to use activity monitors which would certainly be calibrated in whole-body calorimeters in Bangalore. This would amplify the chances of having an objective appraisal of spontaneous changes in physical activity.

12. Concluding remarks

The Sukhatme-Margen hypothesis is useful if it stimulates a more rigorous approach to the sequence of changes which can occur in metabolic efficiency and behavioural changes in physical activity. The wider aspects of analysis which assumes that interindividual differences in energy requirements in practice relate to fluctuations in intraindividual energy requirements, is fallacious on the basis of current evidence, but a detailed analysis of the adaptive process is, I believe, warranted, despite the substantial amount of research on prolonged underfeeding. The behavioural change in physical activity is very poorly documented in either underfeeding or overfeeding and we have, as yet, little insight into the effects of early nutritional deprivation on the adaptive sensitivity and capacity of children and adults.

Contents - Previous - Next