International Journal of Obesity (2008) 32, 1607–1610; doi:10.1038/ijo.2008.147; published online 14 October 2008

Thrifty vs Drifty Gene Theory of Obesity

Evolutionary origins of the obesity epidemic: natural selection of thrifty genes or genetic drift following predation release?

A M Prentice1,2, B J Hennig1,2 and A J Fulford1,2

  1. 1MRC International Nutrition Group, London School of Hygiene and Tropical Medicine, London, UK
  2. 2MRC Keneba, The Gambia


This article challenges Speakman's hypothesis that the modern genetic predisposition to obesity has arisen through random genetic drift in the two million years following predation release. We present evidence in support of the hypothesis that a mixture of famines and seasonal food shortages in the post-agricultural era have exerted natural selection in favour of fat storage; an effect most likely mediated through fertility, rather than viability, selection. We conclude that, far from being time to call off the search, recently developed genetic and bio-informatic methods will soon provide a definitive resolution to this long-standing ‘thrifty gene’ controversy.


thrifty gene, evolution, fertility selection, viability selection, predation release



In 1962, JV Neel published his now famous article entitled ‘Diabetes mellitus: a “thrifty genotype” rendered detrimental by “progress”?’,1 in which he proposed that a genetic tendency to rapidly deposit fat in times of plenty would have been advantageous to the survival of hunter-gatherer populations, especially child-bearing women. Variants of this hypothesis have formed the topic of vigorous debate over the ensuing 45 years. In 1989, Neel rejected his own original thesis that the selective pressure for thrifty genes originated from frequent famines in Palaeolithic times, and reverted to a subthesis in the original article, namely that any selective pressures must have been much more recent.2 Elsewhere, we have developed this idea and argued that famines and seasonal food shortages were rare in hunter-gatherer populations, but became extremely common in agriculturalist societies and, through their influence on fertility, have been the source of major selective pressure on the human genome.3, 4, 5 Speakman has recently argued for the rejection of the idea that thrifty genes may have been under positive selection,6 and subsequently proposed a new hypothesis suggesting that genes predisposing to fat gain have arisen through random drift following ‘predation release’.7 The reader is invited to study Speakman's novel and interesting thesis, summarized in the accompanying article8 before reading these rebuttals to better understand the arguments.

Let us start by stressing that the controversy will not be resolved by this debate. Our intention will be to defend the selective origins of thrifty genes as a still-viable hypothesis rather than to attempt the impossible task of providing a proof. In doing so, we will highlight a number of serious logical inconsistencies in Speakman's critique of the selective origins thesis6, 8 before demonstrating that the assumptions he makes in his genetic drift model7 are, in our view, fatally flawed. We will conclude that, far from being ‘time to call off the search’ as he suggests,6 advances in genetic methodologies mean that now is the ideal time to intensify the search and that such methods should yield a definitive resolution of the controversy within the next few years.

In attempting to refute the foundations of Neel's hypothesis, Speakman raises five key objections, many of which have been raised by others previously,9 including Neel himself.2 Each of these can be rebutted as shown in the following sections.


Mortality in famine is insufficient to lead to genetic selection

We agree with this statement, but propose that it is irrelevant, because it focuses entirely on viability selection, completely ignoring fertility selection. This critical misconception was pointed out by us at the Obesity Society Presidential Debate in 2007 so has no doubt been addressed in the accompanying article8 to which, at the time of writing, we are ‘blinded’. The failure to consider fertility selection is a fundamental flaw in Speakman's arguments and has special relevance to the possible mechanisms of selection, the likely contemporary manifestations of such selection and the time scale over which they are likely to have occurred.

As pointed out by Fisher in his 1930 book on The Genetical Theory of Natural Selection,10 ‘The intensity of selection by differences in fertility … is relatively enormous in comparison to selective intensities to be expected in nature. Fertility selection is sufficient to produce considerable evolutionary changes in relatively short historical periods …’. The human hypothalamic–pituitary–gonadal axis is exquisitely sensitive to maternal energy status mediated by circulating leptin levels.11 Effects are graded,11 as opposed to the on/off effect at very low body fat levels suggested by Speakman.7 They result in almost complete suppression of fertility in conditions of catastrophic famine but, much more importantly, research in The Gambia and Bangladesh,5 and elsewhere,11 shows a 30–50% reduction in conceptions during each annual hungry season (see Figure 1). Hungry seasons have been a norm in most populations, since the dawn of agriculture and, as pointed out by Fisher, differential conception rates among women able or not able to conceive under energetically marginal conditions could have an extremely powerful and rapid effect on selection. It is for this reason that we have hypothesized that if thrifty genes exist, they are more likely to modulate reproductive fitness than survival during energy crises.5 Data from China collected 20 years after the massive famine created by Chairman Mao's ‘Great Leap Forward’ clearly illustrate this point. The age pyramid for Nanquan province shows a virtual absence of individuals aged 19 and 20 years old caused by a severe suppression of fertility. This, and other similar data, shows how regular famines could add to the underlying pattern of seasonal suppression of reproduction and lead to strong selection of women carrying genes that allowed them to continue to reproduce when their competitors had become infertile.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact or the author

Seasonal hunger and its effects on fertility. (a) Seasonality of weight changes in Gambian women caused by the annual hungry and harvest seasons. (b) Seasonality of conceptions in populations suffering from seasonal food shortages in rural Gambia and Bangladesh. Both reproduced with permission of Prentice.4

Full figure and legend (142K)


Historical patterns of famine are incompatible with a selective hypothesis

In this regard, Speakman and ourselves are in part agreement. We agree that famines are unlikely to have exerted selective pressure in the early Paleolithic era, thus refuting Neel's original proposals for the selection of thrifty genes.1 However, we strongly believe that famines and seasonal food shortages have been virtually a norm in sedentary agricultural societies in the past 10000–12000 years and that, through the power of fertility selection, this provides more than ample time for significant natural selection to have occurred.3, 4, 5 In support of this contention, we cite the latest estimates of the dates of origin of resistant variants of G6PD (~2500 years) and TNFSF5 (~6500 years),12 and of lactase persistence (5000–10000 years).13 Note that these are likely to have been under viability selection and that examples mediated through fertility selection, as proposed by us for putative thrifty alleles, could occur over considerably shorter time scales.


In famines few people die of starvation and the burden of mortality in famines affects the wrong individuals

Speakman argues that relatively few people die in famines and that it is the very young and the elderly who are most vulnerable, hence this would have little effect on the gene pool. Once again, this point mistakenly addresses viability, not fertility, selection.


The prevalence of obesity between famines is too low

This point is articulated in Table 1 of Speakman's Cell Metabolism article,7 where he lists the body mass index (BMI, kg/m2) of some modern hunter-gatherer and subsistence agriculturalist populations. First, as argued above, we would not expect hunter-gatherer populations to display a thrifty genotype, as evidence suggests that they have not been subjected to famines. Second, there is now ample verification to show that subsistence agriculturalist populations gain weight rapidly in harvest seasons (see Figure 1) and readily become obese when relieved of food shortages by moving to an urban cash-based economy.14


Criticisms of the ‘genetic drift’ hypothesis

Let us now turn our attention to Speakman's alternative model in which he suggests that early hominids would have been subjected to stabilizing selection for body fatness, with obesity selected against by the risk of predation by large carnivores. He argues that this would have imposed an upper boundary to viable adiposity that would have been eliminated when predation was reduced by the development of social behaviour, weapons and fire, thus allowing BMI to drift upwards without constraint. He then models the effects of random drift and, by making numerous assumptions, has argued that the current BMI distribution in the United States could theoretically be explained by random genetic drift rather than by selection. Space limitations prevent us from tackling all aspects of this argument here, so we will make only two points.

First, Speakman's modelling of the possible consequences of genetic drift following release from a putative constraint to the upper limits of BMI imposed by predation7 is flawed on several grounds. Speakman calculates the distribution of BMI, assuming that dozens of minor mutations on many genes are equivalent to a single major mutation on one gene. This is both incorrect and unnecessary. Suppose N genes are involved. Speakman appears to argue that each mutation independently contributes an 8/N BMI point shift in the upper intervention point. Assuming, as he does, an average rate of 1.5 mutations per gene, the total number of positive mutations will follow a Poisson distribution with a mean of 1.5N. The resulting distribution of the net shift in upper intervention point will therefore have a mean=1.5N × 8/N=12 BMI units and variance=1.5N × (8/N)2=96/N. According to Speakman's assumptions, there will be an identical distribution of mutations with negative effects on upper intervention point. Combining these gives a distribution with a mean of zero and variance=2 × 96/N=192/N. Arguing, as Speakman does,7 that genomes resulting in a net reduction in upper intervention point will be lost by selection, the final distribution will be truncated below zero. The exact form of this final distribution may be approximated by the half-normal distribution with scale parameter √(192/N). Both the mean and standard deviation of this distribution are inversely proportional to √N. Thus, assuming, as Speakman does, that N=1 when it is in fact, say, 100 will inflate the distribution by a factor of 10 and completely invalidate his Figure 3.7 His calculations also posit that every mutation is functional; clearly, an unsustainable assumption given that only 2% of the whole genome comprises protein-coding sequence, not all variation here has functional consequences, and genetic variation in non-coding sequence is much less likely to have an effect at the functional level. Furthermore, his independent/additive model fails to account for population size (which has powerful effects on the rate at which genetic variation will become fixed or lost over time), possible epistasis (that is, gene–gene as well as gene–environment interaction), population bottlenecks and expansions, migration or founder effects, or population subdivision. Nor does he confront the issue of needing to explain why certain genetic traits highly deleterious in the modern environment (type 2 diabetes, polycystic ovarian syndrome and so on) are present at far higher levels than can be explained by random mutation alone.15

Secondly, he argues that the presence of lean people can be accounted for by a model of genetic drift and is incompatible with the notion of selection for thrifty genes. In fact, there are many possible counter-arguments. In modern societies, many people are exerting cognitive restraint to hold their phenotypic BMI below its natural level. There are other possible explanations for polymorphisms yielding varied phenotypes; indeed, the very existence of the term polymorphisms encapsulates this concept. For instance, selective pressures may bring in a succession of small incremental mutations over a long period of time and all these may not have reached fixation; there may have been recent admixtures of populations previously experiencing very different evolutionary (energy supply) histories; and alleles might exist whose advantage is dependent on their frequency in the population. A controversial variant of the last point, suggested by Sewall Wright in 1932,16 is that there may be numerous fitness peaks in an adaptive landscape and these can lead to stabilization of more than one genetic solution to an environmental challenge. Adiposity-related metabolic adaptations are not the only way to survive famine. Elsewhere we have described how behavioural traits can achieve a similar end (for example, those leading to power, wealth and subordination of ones peers; a readiness to migrate when local environmental conditions deteriorate; a highly developed survival instinct including a readiness to steal and so on).5 In relation to fertility, which we argue to be of central importance to this whole debate, there might also be a range of genetically endowed responses to starvation that will carry differential benefit under varying conditions. For example, a genotype that favours continued conception during annual hungry seasons may be beneficial in most years when the hungry season is followed by a successful harvest, but could be catastrophic if the crops fail and the mother is pregnant or lactating when famine hits. Under such conditions those women without this genotype would be favoured. In summary, against a background of constantly variable and unpredictable food supply, overall population survival might have benefited by maintaining a range of genotypes each of which flourished under different environmental conditions. This suggestion is controversial but cannot yet be dismissed.


Future prospects for a resolution of the ‘thrifty gene’ debate

Recent advances in genetic methods make it highly likely that the search for putative thrifty genes, hitherto characterized by speculative hypothesizing, will soon be informed by real data. In the past year, genome-wide association studies have confirmed 11 new gene regions associated with type 2 diabetes,17 and definitive proof has been found for the first of the multigenic contributors to increased fat mass,18, 19 making the prospects for rapid advance in understanding the genetics of complex diseases look encouraging. Furthermore, new bio-informatic techniques are capable of deciphering the time course of recent genetic selection within individual genes and haplotypes, and comparing these between ethnic groups and geographical regions, thus allowing us to progress the process of piecing together the historical influences on selection in the human genome.20, 21, 22, 23 Far from calling off the search for thrifty genes, the moment has never been so favourable, and we predict that ‘thrifty’ variants affecting the sensitivity of the human reproductive axis to energy supply may yet be uncovered.



  1. Neel JV. Diabetes mellitus: a 'thrifty; genotype rendered detrimental by 'progress'. Am J Hum Genet 1962; 14: 353–362. | PubMed | ISI | ChemPort |
  2. Neel JV. Update to 'The Study of Natural Selection in Primitive and Civilized Human Populations'. Human Biol 1989; 61: 811–823.
  3. Prentice AM, Rayco-Solon P, Moore SE. Insights from the developing world: thrifty genotypes and thrifty phenotypes. Proc Nutr Soc 2005; 64: 153–161. | Article | PubMed | ChemPort |
  4. Prentice AM. Starvation in humans: evolutionary background and contemporary implications. Mech Ageing Dev 2005; 126: 976–981. | Article | PubMed |
  5. Prentice AM. Surviving famine. In: Shuckburgh E (ed). Survival: Survival of the Human Race. Cambridge University Press: Cambridge, 2007. pp 146–177.
  6. Speakman JR. Thrifty genes for obesity and the metabolic syndrome—time to call off the search? Diab Vasc Dis Res 2006; 3: 7–11. | Article | PubMed |
  7. Speakman JR. A nonadaptive scenario explaining the genetic predisposition to obesity: the 'predation release' hypothesis. Cell Metab 2007; 6: 5–12. | Article | PubMed | ChemPort |
  8. Speakman JR. Thrifty genes for obesity, an attractive but flawed idea, and an alternative perspective: the 'drifty gene' hypothesis. Int J Obes 2008; advance online 14 October 2008, doi:10.1038/ijo.2008.161.
  9. Joffe B, Zimmet P. The thrifty genotype in type 2 diabetes: an unfinished symphony moving to its finale? Endocrine 1998; 9: 139–141. | Article | PubMed | ChemPort |
  10. Fisher RA. Genetical Theory of Natural Selection. Clarendon Press: Oxford, 1930.
  11. Ellison PT. On Fertile Ground: A Natural History of Reproduction. Harvard University Press: Cambridge, MA, 2001.
  12. Sabeti PC, Reich DE, Higgins JM, Levine HZ, Richter DJ, Schaffner SF et al. Detecting recent positive selection in the human genome from haplotype structure. Nature 2002; 419: 832–837. | Article | PubMed | ISI | ChemPort |
  13. Bersaglieri T, Sabeti PC, Patterson N, Vanderploeg T, Schaffner SF, Drake JA et al. Genetic signatures of strong recent positive selection at the lactase gene. Am J Hum Genet 2004; 74: 1111–1120. | Article | PubMed | ISI | ChemPort |
  14. Prentice AM. The emerging epidemic of obesity in developing countries. Int J Epidemiol 2006; 35: 93–99. | Article | PubMed | ISI |
  15. Diamond J. The double puzzle of diabetes. Nature 2003; 423: 599–602. | Article | PubMed | ISI | ChemPort |
  16. Wright S. The roles of mutation, inbreeding, crossbreeding and selection in evolution. Proc 6th Int. Cong Genet 1932; 1: 356–366.
  17. Frayling TM. Genome-wide association studies provide new insights into type 2 diabetes aetiology. Nat Rev Genet 2007; 8: 657–662. | Article | PubMed | ISI | ChemPort |
  18. Frayling T, Timpson NJ, Weedon MN, Zeggini E, Freathy RM, Lindgren CM et al. A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity. Science 2007; 316: 889–894. | Article | PubMed | ISI | ChemPort |
  19. Loos RJ, Lindgren CM, Li S, Wheeler E, Zhao JH, Prokopenko I et al. Common variants near MC4R are associated with fat mass, weight and risk of obesity. Nat Genet 2008; 40: 768–775. | Article | PubMed | ChemPort |
  20. Sabeti PC, Schaffner SF, Fry B, Lohmueller J, Varilly P, Shamovsky O et al. Positive natural selection in the human lineage. Science 2006; 312: 1614–1620. | Article | PubMed | ISI | ChemPort |
  21. Tang K, Thornton KR, Stoneking M. A new approach for using genome scans to detect recent positive selection in the human genome. PLoS Biol 2007; 5: e171. | Article | PubMed | ChemPort |
  22. Voight BF, Kudaravalli S, Wen X, Pritchard JK. A map of recent positive selection in the human genome. PLoS Biol 2006; 4: e72. | Article | PubMed | ChemPort |
  23. Zhang C, Bailey DK, Awad T, Liu G, Xing G, Cao M et al. A whole genome long-range haplotype (WGLRH) test for detecting imprints of positive selection in human populations. Bioinformatics 2006; 22: 2122–2128. | Article | PubMed | ChemPort |


These links to content published by NPG are automatically generated


Melanocortins and body weight: a tale of two receptors

Nature Genetics News and Views (01 Sep 2000)

Extra navigation