Why We Get Fat

Arthur De Vany June 22, 2005

Contents
1 Introduction 3

2 Homeostatic Theories of Energy Management 2.1 2.2 Set Point Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Thrifty Gene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 4 5

3 Body Composition

4 Fat as an energy adaptation

5 Starvation

6 Non-Linear Dynamics of Body Mass 6.1 6.2 Human Energy Expenditure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8 8 8

7 Statistical Models of Energy Reserves 7.1 7.2

10

The Power Law Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 The Gaussian Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1

8 Positive Energy Balance as an Evolved Adaptation

12

9 Implications and Some Evidence 9.1 9.2 9.3 9.4 9.5 9.6

13

The Human Brain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Set Point Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Body Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Gene Expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Social Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

10 Conclusion

15

11 Appendix A

17

11.1 The Poisson Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Introduction

The simple answer to why we get fat may be that we eat too much. But, as Nesse and Williams would say, that is only the proximate explanation; it may illuminate the underlying physiological mechanisms that result in obesity, but it does not explain why humans are inclined to eat in excess of energy expenditure. An evolutionary explanation requires that we discover the sources of the possibly adaptive value of preferences and behaviors that are biased toward positive energy balance to the human species. To do that I argue that we must move away from models of human feeding and energy management that are based on homeostasis. The phrase energy balance subtly frames the issue of fat around the concept of homeostasis. To say that a person is fat because they overeat and so fail to maintain energy balance is somehow to say that energy balance is a natural or desirable state. A homeostatic theory of appetite and energy management assumes that these systems are integrated in such a way as to maintain energy balance at some level. Is this theory true? What is the appropriate level of energy reserves? Are humans homeostatic energy managers? If they were, then homeostasis would be a natural state. But then how could anyone get fat? If homeostasis is natural, why is energy balance so hard to achieve? It is far from clear that evolution would favor homeostatic mechanisms of energy control in the world of our ancestors. Such a mechanism is costly and may have had little value in an ancestral world of scarce energy sources and high energy expenditure. A mechanism that elicited feeding behavior only when energy resources were depleted and ensured no more than restoration of the depleted energy almost surely would result in starvation in the complex stochastic energy environment of the Paleolithic. Modern obese humans demonstrate that evolved appetite mechanisms are capable of maintaining the body in a pathological state. Pathologically obese individuals do not experience diminished appetite in spite of their very large energy reserves. How could such an appetite mechanism evolve if it were part of a homeostatic system? In this paper I want to argue and demonstrate that homeostatic theories of feeding and energy reserves are incorrect models of human (and mammalian) behavior. Homeostatic theories dominate a good deal of current thinking about obesity, particularly in the public mind and even (according to surveys) the thinking of health professionals. The set point theory of weight control, that makes so many so pessimistic about losing weight, is a well-known variant of homeostatic models. So is the thrifty gene model. I argue that there is no adaptive value in maintaining energy balance in the world of our ancestors and that human metabolism and feeding are not energetically balanced. Homeostasis in energy intake and expenditure and energy stores is not adapted to a complex energy environment. Humans, and other mammals, are designed to feed in excess of energy expenditure. Our evolved instincts are to live in a state of positive energy balance in which we eat too much and exercise too little. This energy bias was adaptive in the world of our ancestors where high energy expenditure and food procurement were intermittent and insecurely linked to one another. I develop the argument by showing that human metabolism is non-linear and that body mass adapts to energy intake and expenditure through a non-linear dynamic that converges to an attractor. This dynamic attractor exhibits some of the features that a set point theory tries to capture, but there is no set point. Then I examine the complexity of the ancestral energy environment using models

of stochastic energy expenditure and intake. I show through statistical and simulation models that the best energy management strategy for ancestral humans was to eat everything in sight and to be as lazy as possible. Energy balance and homeostasis is a fatal strategy. These nonhomeostatic models oer an explanation for the modern dilemma of obesity in humans living in Western societies. We live in a non-complex energy environment of assured and abundant food and earn our living in sedentary jobs that, in eect, pay us not to exercise. In the penultimate section, I discuss some of the implications of the models and the evidence. Then I close with some strategies suggested by the non-homeostatic model for achieving and maintaining a healthful body composition more typical of the human species than the present generation exhibits. I nd very strong results for energy expenditure relative to limiting energy intake. Relatively modest increases in energy expenditure have more power to control body mass than much larger decreases in energy intake. In other words, exercise is more powerful than dieting in bringing about a desirable body composition. Some diets may lead to a dangerous loss of lean body mass, leading to organ collapse and/or heart damage and ventricular brillation.

2
2.1

Homeostatic Theories of Energy Management

Set Point Theory

The homeostatic theory of energy reserves dominates current thinking. The set point theory is a homeostatic theory that proposes that declines in energy resources below their set point produce increased feeding and energy intake until the set point is restored. The set point model dominates current thinking about hunger and eating and weight control of lay persons and even health professionals. In this model, the body seeks to defend its energy stores through compensatory changes in appetite. To remain at a given weight, we must eat what we expend. If the set point theory is right, why is this so hard to do? There is abundant evidence against the set point theory (see the reviews of Brandes, 1977; Friedman, Emmerich, and Gil, 1980; Smith, Gibbs, Strohmayer, and Stokes, 1972; Wilson and Davis, 1977; Woods, 1991). Obese individuals do not suer diminished appetite. Paradoxically, they are often more driven by hunger than people who are closer to a normal body weight. The set point theory fails to account for the rise in obesity most people experience as they age. Are their set points drifting up over time? Is the population of the United States undergoing an upward drift of their set points as the incidence of obesity approaches 50%? What is the adaptive value of a set point energy management system that is capable of putting the body in a pathological state of obesity? The set point theory is also incapable of explaining the wide range of energy reserves that we see among humans. Life can be maintained over a wide range of body weights and what is taken as normal depends on the population. A modern, Western population has a range of body weights far larger than any recorded for traditional hunter gatherers and this is true even when standardized for height by such measures as BMI. Energy balance can occur at many dierent levels of body weight; a vastly over- or under-weight individual can be in energy balance at that weight, but it may be unhealthful. Health requires that energy balance be maintained at a desirable body composition.

It is clear that hunter gatherers and likely our ancestors before them, maintained a dierent body composition from modern humans; they had more lean body mass relative to fat mass. They also were more muscular and so carried more of their energy reserves in protein than in fat.

2.2

The Thrifty Gene

Another theory of energy reserves is the thrifty gene hypothesis. It proposes that evolution would have favored individuals who were ecient in storing energy in a world where food was scarce and high energy expenditure was required to get it. Neel originally proposed the thrifty gene theory to explain why some individuals get fat while others do not. But, in the United States, more than half of us are fat and obesity is on the rise. To explain this pattern, most of us have to have the thrifty gene. That may be, but then what does the theory explain? It is not clear at all that the ability to store energy as fat confers a selective advantage. To do this requires that the body divert the metabolic products of food from lean body mass to fat, sacricing mobility, strength, and protein reserves. Thus, a thrifty genotype who stores more energy as fat would be physically smaller than non-thrifty genotypes, with a smaller muscle and organ mass relative to body fat. What advantages could this altered body composition confer in survival and reproduction? How eective would a smaller ancestor with high body fat relative to lean body mass be in the competition for resources and mates during the Paleolithic against larger and more muscular, leaner individuals? I fail to see how a thrifty gene that favors the deposition of fat would confer a selective advantage against non-thrifty genotypes. The evidence suggests that ancient as well as 20th century hunter gatherers carried more muscle and less fat than modern people. Moreover, the thrifty gene theory leaves open the question of homeostasis: does the thrifty gene set a high set point for energy reserves or does the gene enhance energy storage? In the rst case, energy balance will be maintained at an altered body composition, in the second, energy balance and body composition may drift. But, if energy balance drifts, then the theory fails to explain body composition. Clearly, with a thrifty gene or not, it must be possible for individuals to ingest more energy than they expend if they are to become obese. The thrifty gene hypothesis is an unsatisfying explanation (and may never have been oered for more than an explanation of why some get fat). We must make a distinction between an explanation of why an individual gets fat and another does not, versus an explanation of why all members of a species are vulnerable to obesity.1 Explaining why some individuals get fat, say the Pima, and explaining why obesity is so common are separate problems. It is the latter that I am addressing.

Body Composition

Fossil analysis indicates our species was derived from a tropical primate ancestor. Humans are not physiologically adapted to severe cold. Even Eskimos are lean, on the order of 11% fat living the
1 This paraphrases Nesse and Williams, p 17 where the discussion is about disease, but the distinction applies equally well here.

old ways. The Eskimo, as other humans do and have done for all time, use technology to cope with cold. Other animals do adapt their fat content to climate: whales carry up to 35% fat as insulation against cold, and the body fat composition of animals declines from the cold poles to the equator. Cold-stressed humans, even those in populations known to have survived cold climates for millennia, survive through energy-consuming heat production rather than through the use of insulation. Metabolic heat produced through muscular activity, shivering or thermogenesis in brown fat compensates for heat loss to the environment. Even though humans are among the fattest of mammals, human response to cold stress does not rely on the build-up of fat. In fact, heat stress also increases the individuals energy needs. And human populations living in hot climates have as much subcutaneous fat as, or more than, the Eskimo.

Fat as an energy adaptation

If fat is not a cold adaptation then it is likely to be an energy adaptation; that both heat and cold increase energy demands and makes the energy component of fat common to both. Thus, its role as an insulator is secondary to its role as a source of energy. Such reserves can be drawn upon during times of food shortage and cold stress to support thermogenesis. Human fat depots are located at various sites around to body. Their distribution seems poorly designed for insulation and is more likely a consequence of the ease of carrying fat as an energy store and as a shock absorber for the organs. Fat reserves are mobilized by epinephrine which splits the triglycertriglyceridesides into free fatty acids and glycerol. The FFA are then split into acetyl-coenzyme A, which enters the carboxylic acid cycle. During cold stress, lipoprotein lipase activity decreases in white adipose tissue and increases in brown adipose tissue. We were omnivorous from the beginning. Dentition suited to carnivory and plant eating exists in fossils over 2.5 million years old. Human ancestors had the ability to exploit a range of foods, animal and vegetable matter. Dietary diversity permits adaptation to changing climate and food sources. Dependance on a single source is risky. Storage of nutrient reserves (energy and vitamins and minerals) in fat is an essential element of that adaptation. The human brain has energy demands that are unlike those experienced by any other species. The brain consumes from one quarter to one third of the bodys basal energy demands. This is on the order of 400-600 kcal/day. Since the brain relies on glucose, this translates into a demand for 100 to 145 g of glucose/day. Coupled with this energy requirement is a high oxygen demand. The brain uses about 45% of the oxygen we breath and about two thirds of the bodys circulating glucose. These are demands which must be met or irreversible damage will occur. Contemporary brain size is from 1200 to 2000 g, the male brain being larger and requiring more calories. A progressive increase in brain size occurred during pre-hominid and hominid evolution. Australopithecus had a brain of 450-500 cc 2 million years ago; Homo erectus had a brain 1000-1100 ccs big 1.5-0.5 million years ago. Coinciding with the growth in brain size, and energy requirements, is evidence of increasing technological sophistication in tools whose purpose is directly related to hunting and the use of animal sources of food. This is no coincidence, for the larger brains energy requirements could only be met with nutritionally dense animal food. Tool making, large brains, and meat eating coevolved together.

Starvation

Accompanying this coevolution of nutrition and large brains was the evolution of fats adaptive value as an energy store. A fat reserve gives one a longer term of survival during times of nutrition stress. But, weight loss during starvation (or a starvation kind of diet) is more than simply living o ones fat. There are energy and protein requirements to be met, as well as the need for essential fatty acids. Vitamins, minerals and water are also requirements. To satisfy all these demands during food stress the body uses several lines of defense. First, energy demand is met by freeing glucose from liver glycogen. This lasts a few hours. Amino acids released from tissue then come into play. This is accompanied by a loss of minerals. To prevent electrolytic imbalance, there is a shedding of water, the main cause of weight loss during the rst 5 days of a fast. Then fat losses begin. This too is non-linear. From the fth to the ninth day of fast, about 4% of body weight is lost most of which is fat. The rate of loss is about 500 g a day, but this drops to 100 g/day by the 25th day of the fast. At this lower rate, a person can last a long time with no food. But, not only is fat lost. Lean tissue and organ mass also are consumed. Even heart muscle is sacriced. A 20-year old girl who lost 128 lb. in 30 weeks consumed half her bodys lean tissue mass, including heart muscle. She died of ventricular brillation. The loss in lean tissue is required to supply the bodys continuing needs for amino acids. The muscle is a rich store of amino acids. When given up, these amino acids are not replaced unless the fast ends and protein is eaten. The released amino acids are consumed in gluconeogenesis and in the synthesis of antibodies, essential to the immune system. Exposure to infection increases the demand for amino acids to synthesize antibodies. As mammals, humans are continuous metabolizers and discontinuous feeders. This presents an inventory problem for energy reserves. Energy must be stored between meals and, given the relatively limited capacity for gluconeogenesis, mechanisms must be in place to preserve glucose for the high energy human brain. We may owe our large brains to our ability to get fat. Carnivores obtain long-chain fatty acids, essential to big brains, from other animals (Crawford and Marsh). Humans have large brains and small stomachs. We clearly are not herbivores, though plants occupy an essential part of the human diet. Our metabolic rate is very high, because of our large brains, so high energy foods are essential, along with the long-chain fatty acids. We get them from animals, whose availability is seasonal and episodic. So, early humans embarked on a risky strategy for survival. We must therefore be able to store energy well, better than wild animals, who have only 4% body fat and better than chimpanzees, the only primate other than humans that tends to get fat. Humans are, therefore, at the high end of the scale in their ability to store fat. A large-brained, small-stomached omnivore with long maturation time living on scarce, high value nutrients and seasonal plant foods of varying quality faces a serious stochastic energy balance problem.

Non-Linear Dynamics of Body Mass

In the following model, I discard the concept of homeostasis and argue that humans are far-fromequilibrium, open energy systems. The conditions of maintaining life are, therefore, dierent from those given by a linear, stable system.

6.1

Human Energy Expenditure

Farmers in Nepal expend from 1.89 to 2.01 times their basal metabolic rate (METS) in a day of activity. A worker spends about 1.7 METS. The Ache expend about 1.9. Oce workers expend about 1.3. Lumberjacks expend 6-8 METS. The highest sustained metabolic rates were recorded in the Tour de France where, over a 22 day period, the cyclists maintained rates that were 4.3 times their basal rates. Mice, bees and other animals do not generally exceed 7 times basal metabolic rates on a sustained basis. All these eld metabolic rates are well below the maximum that could be attained. Theoretical limit may be the inter-species limit of 1.7M J/day/kg 0.72 . Competitors in the Tour de France achieved over a 3 week period a daily energy expenditure of 1.6 1.8M J/day/kg 0.72 . Higher rates seem to threaten survival and reproduction. During the same period, the Tour de France competitors consumed an average of 24.5 MJ/day and expended 29.4 MJ/day. They lost about 1.5 kg of lean body mass and about 2.3 kg of fat mass. Clearly, humans can expend energy well in excess of intake over a sustained period of time. Of particular interest is the adjustment of body mass to the sustained negative energy balance. This loss is not atypical of dieters, though dieters do tend to lose more lean body mass because they are less active than these competitive cyclers.

6.2

Model

Based on these considerations, I model body mass change in the following way. Let w be the change in body weight or mass in kg. Let ei be energy intake and let ee be energy expenditure. To complete the notation for the variables in the model, let m be the metabolic rate as a proportion of the maximum sustainable rate of 1.6 1.8M J/day/kg indicated in the data presented above. Estimates of the conversion of energy into body mass place the eciency at 0.75. I include the thermic eect of food as a loss of ei0.1 . The equation for change in body mass then is

w = 0.75 (ei ei0.1 m w0.72 )

(1)

In this model, energy balance is achieved when the change in body mass is zero. Body mass adjusts to a positive balance until body mass raises energy expenditure to equal intake. The response to a negative energy balance is similar, though in this instance body mass declines until energy expenditure matches intake. The solution for steady state body weight, given energy intake and 8

expenditures of ei and m is the value w such that w = 0. Because w is a concave and continuous function in w a solution will exist. The solution will depend on the parameters the human subject sets; these are ei and m, representing energy intake and expenditure in multiples of the maximum sustainable metabolic rate. For energy intakes of 5, 10, 15, 20 and 25 MJ/day and a metabolic rate of 0.5 of the maximal rate of 1.7MJ, the corresponding steady state body masses at which expenditure equals intake are 16.87, 53.19, 99.17, 152.38, and 211.56. The caloric values of these intakes are from just over 1000 to about 6,000 and the body weights, in pounds, range from about 35 to over 440. On the expenditure side, for an intake of 12 MJ/day and metabolic rates of 0.2, 0.4, 0.6, 0.8, 1, the corresponding steady state body masses are 252.05, 96.25, 54.80, 36.75, 26.95. Modest increases in energy expenditure bring large reductions in body mass. The dynamics of the process are very revealing. The solution for body mass now becomes a xed point problem; starting at a mass of m0 the dynamics are a mapping to m1 . We seek the solution as a xed point w of the dynamic mapping M (wt ) wt+1 ; this is just the recursive dynamic wt+1 = wt + w. Since the energetics are non-linear and concave, this will be a contraction mapping and a xed point exists and is stable. The gain segment of the mapping is shown in Figure 1. The loss segment is shown in Figure 2. It has a similar curved shape. The path to an equilibrium body weight has the distinctive curve typical of weight gain and loss experienced by most people: an initially rapid gain or loss followed by a slowing and attening. This path is suggestive of a set point in that the rate of gain or loss eventually diminishes with further departures from the initial weight. But, what is really going on is that the path to the equilibrium weight to which the process converges is non-linear. There is no set point, but there is a non-linear convergence to one of a multiplicity of equilibria, each of which is individually consistent with a particular pair of values of energy intake and expenditure (ei, m). Several propositions can be shown to follow from the model. The equilibrium is independent of initial body weight. The control parameters for each individual are ei and m; the remaining parameters are physiologically determined. Hence, a dieter who is undergoing acute weight loss will not have reached equilibrium and may experience a transient reduction in metabolic rate because of diminished lean body mass along with the fat. Thus, energy expenditure will diminish even if activity is maintained at the rate set per unit of lean body mass. The energy decit associated with the initial conditions will diminish over the course of the decline in weight, and eventually reach zero, where the equilibrium weight will be established. Unless the subject continues to operate at the same energy expenditure per unit of lean body mass and while maintaining the same energy intake, weight will not stabilize. The weight regain that will be experienced if the subject returns to the energy imbalance that led to the over weight condition will be rapid initially and non-linear also. Thus, an acute diet or exercise program that leads to a loss in weight will be followed by a rapid regain once the energy decit is eliminated and a surplus corresponding to the original conditions is restored. Such an individual, cycling through these acute stages, will seem to experience a return to a set point.

Further implications of the model and the consistency with the evidence of these results will be discussed later. Now we turn to the adaptive value of fat, that is to the evolutionary question: what is fat for?

Statistical Models of Energy Reserves

Humans are capable of expending levels of energy that exceed energy intake on a continuing basis. To do this, they rely on energy stores in muscle and liver glycogen, on gluconeogenesis of lean body protein, and on fat. By far the largest reserve of energy is contained in body fat, though this clearly varies with the body composition of the individual. It has been estimated that even relatively lean individuals carry enough body fat to support their basal metabolism for about 30 days. Less lean, but still not obese individuals may carry between 60 and 90 days worth of fat. Obese individuals carry enough fat to support their basal metabolism for one year. People living at a positive energy balance of as little as 3% will gain more than 40 pounds between the ages of 35 and 55, and this is roughly the pattern in the United States. There is scant evidence to suggest ancient humans were obese and available evidence of naturally living hunter gatherers indicates they are uniformly lean, so we have a puzzle. My attempt to solve it is to look at the energy management problems humans may have faced in the roughly 3 million years that preceded the advent of agriculture. I develop and examine stochastic models of energy expenditure and intake. I assume the form of statistical distributions of food encounters and energy expenditure and then examine the probability of survival under dierent energy balance strategies. I examine models under three forms of statistical distributions. The rst model assumes a Poisson distribution of prey encounters. Since the equilibrium of this model is death, it does little more than to illustrate the high stakes of the foraging life. I report the model briey in Appendix A. The second assumes a power law distribution of energy expenditure and intake. The third is a variant of the second in which the distributions are gaussian. All the models show that a positive energy balance strategy gives the highest probability of reproduction and survival. It would be more informative to follow the energy state of the forager and to use more realistic statistical models. We turn to that task.

7.1

The Power Law Model

The energy environment of the Paleolithic was complex. There were intermittent periods of high energy expenditure mixed with expenditures on many scales, with a high natural rate of expenditure relative to Western living. A human forager would have frequent encounters of low energy density plant food and less frequent encounters with high density animal sources. The possibility of encountering extremely large prey would make the distribution of energy content of encounters highly skewed. A Rhinoceros or a Mammoth presents such a mass of energy that one or more such encounters dramatically alters the average energy encounter rate. I assume that energy intake and

10

expenditure follow power law distributions. I adopt a Pareto distribution as the model. A Pareto distribution is consistent with the energy expenditure data discussed above and would also appear to be a good approximation to the yields a forager would obtain. The possibility of extremely large prey, such as a Rhinoceros or Mammoth, is well captured by the long tail of the Paretian distribution which places more probability density on large events than would the Gaussian distribution. The same may be said on the expenditure side. The Paretian distribution, with its long tails, provides a model of high frequency of low expenditures with rare, but large energy expenditures of the sort one would expect of a forager trekking long distances, hunting dangerous prey, and carrying heavy loads back to camp when the hunt is successful. The Paretian energy intake distribution contains two parameters, a minimum and a tail weight. I choose these to reect human energy metabolism. I set the minimum yield from foraging at 0.5, which is scaled relative to basal metabolism. So, the value reects that the minimum yield from a days worth of foraging is about one half of basal energy expenditure. Corresponding to the minimum energy intake, I set the minimum energy expenditure at 1.54, which reects an expenditure of 1.54 times the basal rate. This is the minimum for an active forager. The tail weight of the distribution of energy intake is set to equal 1.2. The tail weight of the expenditure distribution is set a 3.8. The parameters of the expenditure distribution are chosen to give a mean energy expenditure consistent the evidence on hunter gathers. These parameters give a mean HG MET of 1.909, right about at what Leonard and Robertson estimate for the Ache. With these parameter values, the means for intake and expenditure are 3 and 2.09, reecting a kind of statistical abundance of food relative to energy expenditure. The variances of intake and expenditure are and 0.6386. The innite variance of the Pareto distribution for the tail weight 1.2 simply reects the high variance nature of energy expenditures of foraging. Innite variance is not surprising here, it simply means that the probability in the upper tail declines less rapidly than x2 rises. This is perfectly consistent with the data gained monitoring the movements of wild sh and with estimates of the fractal dimension of the landscape over which our forager must travel. A picture of the cumulative probability distributions is shown in Figure 3. With these distributions, the probability of energy intake less than expenditure is Pn = P [ei ee < 0] = (1 pdf (ei > h))(pdf (ee > h). The probability of negative energy balance can be obtained by numerically integrating this expression. The value quickly converges to 0.3437. The probability that depletion would go on for a sucient number of days to deplete reserves of R can be approximated by the probability of a period of negative balances in succession sucient to induce starvation. This is S , where S = B/r is the number of days that would deplete an energy balance of B at the rate Pn of r METS per day. Clearly, the chances of survival are greater the larger is B . To maximize the chances of survival, a forager in this environment must carry the maximal energy balance. Hence, it pays to have no upper bound on the energy the human body can store. Episodes of negative energy balance will be sucient to maintain body fat in a physiological range, unless it becomes fully depleted and death ensues. I have simulated the energy reserves under the conditions of the model to get some idea of the variation that occurs and to gain some appreciation of the likelihood of death. These experiments are shown in Figures 5, 6, 4. The gures show that energy reserves are volatile, but they do not progressively become larger,

11

even though the forager has been placed in an environment of abundant food. The reason is that energy reserves are periodically drained through high expenditure, low intake, or both. The large excursions through which energy balances move are pretty dramatic and indicate periods of substantial withdrawals and gains. What if the forager could take in no more than a portion of the food captured? Suppose the upper limit that could be consumed in a time period is three times the mean energy expenditure. That means that the forager could eat no more than three times average energy expenditure no matter how large the take in a given episode. Such a limit makes survival more precarious, which is evident from the calculation above and in the simulation of energy balance with an upper barrier in Figure 7 Energy balance is, on average, lower, makes smaller gains, and suers losses that may substantially exceed the gains. Death by starvation becomes more probable. It seems unlikely that limits stemming from energy balance considerations would have been adaptive in this environment. The foragers best strategy, in the absence of food storage, is to eat to physiological limits whenever possible. If these limits can be extended by storing food, so much the better.

7.2

The Gaussian Model

The same kind of analysis can be done with gaussian distributions of energy intake and expenditure. The results strongly suggest that it is the high variance environment, not the mean energy intake and expenditure that controlled body weight of pre-agricultural humans. When gaussian models are used with mean values close to those of the Paretian distributions, the results change dramatically. The gaussian distributions have less variance than the Paretian distributions, and the probabilities of large excursions in energy balance become small. The result is a diminished probability of death by starvation and a much larger mean energy balance. This is true even with a much smaller bias of mean energy intake over mean expenditure than was true with the Paretian distributions. Survival is still far from assured, though death has a diminished probability relative to the Paretian world. The results show that our forager is fatter in a Gaussian world because energy reserves are drained less often and the size of the drain is smaller each time one occurs. Most interesting is the result that shows the forager gaining weight apparently without limit, something that did not occur in the Paretian world. The low variance environment promotes weight gain with even a small tendency to eat more than energy expenditure on those occasions when one does eat. One can suggest that the relatively assured supply of food and low variance energy expenditure are signicant factors in modern obesity.

Positive Energy Balance as an Evolved Adaptation

The statistical models and simulations demonstrate the demanding energy management problem posed by the human adaptation. The solution is quite clear: eating to restore energy reserves is a fatal strategyset point humans are dead. In their place are humans who posses the ability to

12

eat well in excess of energy expenditures, and to do it on an extended basis. Such was a necessary adaptation to uncertain energy expenditure and intake, which likely were distributed in a manner that spelled intermittent long spells of low intake and high expenditure. This is the prominent feature of the power law or Paretian distribution of energy expenditure and encounter rates with game. Enduring these spells required the ability to store energy in substantial amounts. More importantly, it requires an uncoupling of appetite from energy expenditure. This means that in relatively food replete conditions and sedentary activity levels, humans must have the appetite to consume foods well in excess of energy expenditure. Given the intermittent, rare, but sometimes extreme episodes of large energy drains, an appetite mechanism that limited food intake during times of energy surplus would not be sucient to sustain the individual during these periods of large energy decit. The risky strategy of the human adaptation fundamentally altered the nutritional requirements of humans, and especially of modern humans, relative to other animals. Humans were programmed to eat every thing in sight and lost the close feedback between energy expenditure and intake. They are programmed to balance energy over a longer time scale than other animals whose energy requirements are lower, less intense, and whose food sources are readily available and not fugitive. Insulin resistance is part of this adaptation. Insulin resistance spares scarce glucose for the brain. Following vigorous exercise, there is a transitory increase in insulin sensitivity and a decrease in the utilization of protein that is used for gluconeogenesis. These changes serve to replenish muscle glycogen and to conserve on protein to make it available for muscle repair and rebuilding. Glucose is also spared because GH lipolyzes adipose tissue so that free fatty acids serve as the energy substrate to support recovery. Fat is there precisely to support this regenerative process and to conserve glucose for the brain and for glycogen synthesis. For this to work, muscle has to be relatively more insulin sensitive than fat so that the energy available is redirected to brain and muscle tissue rather than into fat. There would have been a constant cycling of energy into and out of fat cells during the Paleolithic. This is something that does not occur among modern humans if they are physically inactive.

9
9.1

Implications and Some Evidence

The Human Brain

It is because we have big brains that we get fat. Insulin resistance is part of the equation for it conserves on glucose to preserve the brain, the essential instrument for our survival. In addition, a wide foraging range is required to exploit this strategy of fugitive high quality nutrients and omnivory. This adds an energy component of high activity per calorie and gram of protein gained for humans, which again is above that of many other animals.

13

9.2

Set Point Theory

Obese persons generally regain lost weight, suggesting that adaptive metabolic changes favor return to a preset weight. But experiments (Weinsier RL, Nagy TR, Hunter GR, Darnell BE, Hensrud DD, Weiss HL, Do adaptive changes in metabolic rate favor weight regain in weight-reduced individuals? An examination of the set-point theory. ) show that body composition adjusted RMR falls only during the acute phase of weight loss. Once body mass reached a steady state, adjusted RMR returned to the base state, before the weight loss. This is just as our model predicts and it shows that direct experiments reject the set point theory, which requires that adjusted RMR must be lower in the new steady state in order to promote a regain to restore the set point. As the model also shows, subjects who regain weight are those who fail to establish energy balance after the weight loss. They regain lost weight because they return to the energy intake and expenditure that determined their steady-state weight they had before they began the weight loss program. This gives the misleading impression that weight-reduced persons are energy conservative and predisposed to regain. What they are predisposed to are the habits that caused them to become overweight and when they return to them following a weight loss, they return to the weight consistent with those habits.

9.3

Exercise

It is not overeating, but under-activity that causes obesity. We are paid not to exercise now. There is persuasive evidence that much obesity is due to under-exercising rather than overeating. In a series of randomized, controlled trials it was found that sedentary men who take up jogging lose body fat in proportion to miles run, increase their energy intake, and improve their lipoprotein pattern. In a 1-year comparison of fat loss by dieting vs. fat loss by exercise without dieting, both methods were found to be eective in moderately overweight men, and both approaches raised plasma HDL cholesterol. It has also been demonstrated in overweight men and women losing weight on a prudent diet (low fat, low cholesterol) that adding exercise to energy restriction further increased loss of body fat and reduced waist-to-hip girth ratio, especially in men. Risk of coronary heart disease was also substantially further reduced by addition of exercise, in both sexes. These studies suggest that regular exercise is a valuable addition to dietary change for weight control and reduction of risk of chronic disease in people of all ages. The energy demands of physical activity must be suciently intense to trigger these changes. Few modern humans engage in activity vigorous enough to do this. And there must be enough muscle tissue relative to fat tissue for the muscle to generate a suciently high energy demand to draw down on fat reserves. If there is a high volume of fat relative to muscle tissue, then, even though muscle is more insulin sensitive than fat, the fat volume will contain more receptors than the muscle tissue. Consequently, energy will ow to fat on balance. The body composition of modern humans makes it easier for them to get fatter. They have so much fat that their fat out competes their muscle (and organs) for energy. In addition, they are incapable of the vigorous activity that would make inroads into their fat inventory.

14

9.4

Body Composition

Modern humans have a body composition that is novel from the long perspective of the evolution of the species. We have relatively more body fat relative to muscle than all of the hunter gatherer populations that have been studied. Even the chubby looking Eskimo has only 10% body fat, so long as the subject lives in the traditional ways. Skeletal remains of pre-agricultural humans show robust skeletons and strong attachment points on limbs that anchored powerful muscles. Altered body composition is an adaptation to energy surplus. The dynamic model developed above only treats total body mass, but it is clear that energy surplus increases fat mass relatively more than lean muscle mass. The growth in body mass is the bodys attempt to achieve energy balance, even at what may be a dramatically elevated energy surplus.

9.5

Gene Expression

Genetic obesity begins at a young age, when gene expression is triggered by low energy ux, poor nutrition, and excessive sugar and fat. This gene expression hypothesis may be the thrifty gene or metabolic syndrome or some more complex process. We all (or most of us) have it to some degree, while the thrifty gene is hypothesized to be carried by only some of us. The gene expression theory of obesity is dierent from the energy balance theory. It is a theory of body composition. It goes something like: energy ow over metabolic pathways alters gene expression and body composition. The ux on each pathway and the intermittent pattern turn on gene expression. So, you dont have to have a thrifty gene to get fat, you can just turn on genes that favor fat deposition through your activities and eating patterns. By the time we are older, the pathways and gene expression already developed favor obesity. So, diets dont work and this is well-established and powerful evidence.

9.6

Social Organization

A social organization that facilitated the sharing and storage of food would clearly have been a useful adaptation given the relatively poor prospects of survival of a forager in the conditions we have modeled. Moreover, these innovations would have contributed to the evolution of a smaller stomach and release resources for metabolically expensive brain tissue. The evidence that social organization seems to have coevolved with larger brains is fully consistent with this.

10

Conclusion

Even obligate carnivores do not have the intense energy demands of the human brain and do not get fat; their body fat is on the order of 4%. Unlike wild animals and game, humans can and do 15

greatly exceed 4% body fat. Our other big-brained relative, chimpanzees also are capable of piling on excess fat and on average carry more body fat than other animals. A natural level of body fat is something like 11% among hunter gatherers. I have shown that the eat everything in sight strategy is optimal for an ancestral forager. It also is a good strategy to minimize energy expenditure relative to return. These pose a dilemma for modern humans; we seem designed to be lazy overeaters. Clearly, an uncoupling of intake and expenditure is an essential adaptation. We have to be capable of overeating, even when sedentary, in order to overcome the large energy drains that would have characterized the uncertain energy environment of the Paleolithic hunter-gatherer. Another point that comes through clearly in the models is that energy balance is something that would be achieved over relatively long time intervals for a foraging Paleolithic human. They would have been in energy decit as much as one third of the time and would experience some rather painfully long periods of hunger stress. But, given their intake and expenditure strategies, few would starve to death unless there were substantial shifts in the plant and prey distributions. The modern idea that we should be in energy balance on an almost daily basis is not a natural condition and is dependent on a reliable supply of food. Almost surely, our ancestors spent intermittent and sometimes demanding episodes of energy decit. The cycling of energy balance through bursts of expenditure and intervals of hunger is a powerful signal that alters gene expression. Caloric deprivation studies seem to show chronic and acute energy decit extends life and improves health indicators such as insulin sensitivity. A related point is that increasing energy expenditure through activity or exercise is more powerful in attaining a healthy body composition and mass than is decreasing energy intake. And extreme diets can be very dangerous. We may owe our brains to our ability to get fat. Now, we just have to use them to overcome our evolved instincts to live in a modern world of energy surplus.

16

11
11.1

Appendix A
The Poisson Model

Assume that food encounters are discrete and follow a Poisson process. We will consider four discrete energy states for the human forager who eats the food encountered in the following fashion. If no prey is encountered, the forager goes hungry and one unit of energy is drained from reserves. If one prey is encountered, the prey is eaten and energy taken in just covered the unit expended in foraging. If two prey are encountered, they both are eaten and energy balance goes up one unit. If more than two prey are encountered, only three are eaten and energy balance goes up to two. Thus, the energy states are 0, 1, 2, 3. From the Poisson assumption, we derive the transition matrix between states. The matrix of discrete transition probabilities is 1 0 0 0 d(0) d(1) d(2) 1 (d(0) + d(1)) 0 0 d(1) d(2) 0 0 d(0) d(1)

m=

This simple model can be solved for the steady state probabilities of the states. Letting the vector of steady state probabilities be ps, the solution satises m ps = ps and p1 + p2 + p3 + p4 = 1. The steady state solution for any values of the parameters of the Poisson distribution is the probability vector for the states of (1, 0, 0, 0). That is to say, the probability of death by starvation is one in the long run. The forager should eat everything she encounters. Preservation and storage would help as this would let the forager carry over encounters of prey larger than he can consume.

17

Figure 1: Non-linear Loss

74

73.5

73

72.5

72

71.5

50

100

150

200

18

Figure 2: Non-linear Gain

99

98

97

96

50

100

150

200

19

Figure 3: CDFs of energy expenditure and intake

0.8

0.6

0.4

0.2

10

-0.2

20

Figure 4: Energy Reserves: Run 1

200

180

160

140

200

400

600

800

1000

21

Figure 5: Energy Reserves: Run 2

180

170

160

150

140

130

120

200

400

600

800

1000

22

Figure 6: Energy Reserves: Run 4

80

60

40

20

200

400

600

800

1000

23

Figure 7: Energy Reserves with an Upper Barrier

35

30

25

20

15

10 20 40 60 80 100

24