The provisioned primate: patterns of obesity across lemurs, monkeys, apes and humans

Abstract

Non-human primates are potentially informative but underutilized species for investigating obesity. I examined patterns of obesity across the Primate order, calculating the ratio of body mass in captivity to that in the wild. This index, relative body mass, for n = 40 non-human primates (mean ± s.d.: females: 1.28 ± 0.30, range 0.67–1.78, males: 1.24 ± 0.28, range 0.70–1.97) overlapped with a reference value for humans (women: 1.52, men: 1.44). Among non-human primates, relative body mass was unrelated to dietary niche, and was marginally greater among female cohorts of terrestrial species. Males and females had similar relative body masses, but species with greater sexual size dimorphism (male/female mass) in wild populations had comparatively larger female body mass in captivity. Provisioned populations in wild and free-ranging settings had similar relative body mass to those in research facilities and zoos. Compared to the wild, captive diets are unlikely to be low in protein or fat, or high in carbohydrate, suggesting these macronutrients are not driving overeating in captive populations. Several primate species, including chimpanzees, a sister-species to humans, had relative body masses similar to humans. Humans are not unique in the propensity to overweight and obesity.

This article is part of a discussion meeting issue ‘Causes of obesity: theories, conjectures and evidence (Part II)’.

1. Introduction

Obesity presents a global public health challenge for both children and adults [1]. The causes of obesity are complex, multifactorial and a focus of continued and intense debate. To date, the vast majority of research on obesity has either examined humans directly or has tested proposed mechanisms using rodent models. Non-human primates are more closely related to humans and may therefore provide a more salient biological model. However, with some notable exceptions [2–6], far less work has been done with non-human primates in obesity research. Further, the work done to date has focused almost entirely on two genera, macaques (Macaca spp.) and baboons (Papio spp.), with little investigation of other species, including our closest relatives, the great apes. In this paper, I examine patterns of obesity across the full expanse of the Primate order to test and refine current models of obesity in humans.

The exclusion of most primate species from obesity research reflects the focus on research laboratory populations in biomedical studies. There are over 70 000 primates in research facilities in the USA alone, but the diversity of species is low, with macaques accounting for the large majority [7]. Zoos and sanctuaries provide a much broader range of species, with standards of care that prioritize the animal's physiological and psychological well-being. Research in these settings can present logistical challenges, as protocols must be non-invasive or minimally invasive. However, work in zoos and sanctuaries is growing [8–10], demonstrating the research potential of these facilities and providing a new model for research that puts the animal's well-being first.

Obesity and cardiometabolic health are common concerns for captive primate populations [8,11–14]. Many species are known to develop obesity in zoo and laboratory settings, where overweight and obesity are typically tracked by body weight and visual determinations of body condition by veterinary staff. As with humans, the causes of obesity in captive non-human primates are not well understood. The magnitude of obesity also tends to increase with age, as in humans in industrialized countries [8,12]. Available measures of daily energy suggest reduced activity expenditure is unlikely to be a major cause of obesity in captivity. In the one species of primate (Lemur catta) with total daily energy expenditure measured in both the wild and captivity, daily expenditures were greater in captivity despite similar mean body mass in both cohorts (wild: 0.61 MJ d⁻¹, 2.2 kg; captive: 0.91 MJ d⁻¹, 2.2 kg) [15]. Similarly, analyses comparing species' expenditures in the wild to those of other species measured in captivity have shown no difference in daily energy expenditure in analyses controlling for body mass [15]. Instead, as with humans, the primary driver of overweight and obesity in captive primates appears to be overconsumption.

The diets of captive primates differ substantially from those in the wild, which may contribute to increased energy intake. Highly processed food (chow) accounts for the majority of calories in captive diets, with the remainder from domesticated fruits and vegetables [16,17]. Consequently, diets in captivity are more highly processed, more energy dense, have lower moisture content, and contain less fibre than diets in the wild [17–19], factors that have been associated with overconsumption and obesity in humans [20,21].

Chow-based diets may also contain more protein and fat, and conversely less carbohydrate, than wild diets. Chow formulae differ, but generally obtain roughly 15–25% of their energy from protein, 10–20% from fat and the remainder from carbohydrate. By comparison, protein generally accounts for 10–15% of dietary energy for primates in the wild [18,19,22–24]. Fat consumption in the wild appears to be more variable, for example ranging from less than 5% of energy for wild gorillas and chimpanzees to approximately 16% of energy for wild orangutans [19,23,25]. These macronutrient differences may in turn affect weight gain in captivity. The ‘protein leverage' hypothesis proposes that increased dietary protein should curb overconsumption as the organism reaches its target protein intake earlier, with less total energy consumed [26,27]. The ‘carbohydrate-insulin model' predicts that consuming a smaller proportion of daily energy as carbohydrate should reduce insulin secretion and thereby protect against fat deposition and unhealthy weight gain [28]. Both hypotheses would predict that non-human primates in captivity should be protected against obesity, since their diets in captivity are generally higher in protein and lower in carbohydrate than in the wild. In particular, frugivorous primates often eat comparatively less protein and more carbohydrate in the wild relative to folivores [29]. Therefore, frugivorous primates might be evolved to target lower protein intake and thus be less susceptible to overconsumption in captivity and obesity than folivorous species.

The environmental and psychosocial context of captivity may also affect consumption. Primates in captivity, like humans in the industrialized world, live in an environment where food is readily available and requires little effort to find and acquire. Food insecurity, which is thought to promote overconsumption in humans and other species [30], is therefore reduced. Other psychosocial stressors are more difficult to assess. Predation is not a concern in captivity, and staff generally work to minimize conflict within captive populations, but social stresses, boredom and in some contexts social isolation may still persist. In humans, chronic psychosocial stresses are associated with overconsumption and obesity [31,32].

Susceptibility to obesity may also be influenced by evolutionary history. Neel, in his classic the ‘thrifty genotype' hypothesis, proposed that a history of ‘feast and famine' in food availability among humans' hunter–gatherer ancestors selected for alleles that increase insulin resistance and promote fat accumulation [33,34]. Similarly, Mellor et al. [12] in a study of obesity in 13 lemur species (six genera) reported a trend suggesting that species adapted to more unpredictable foods could be prone to being overweight in captivity. Ripe fruit is generally less available and undergoes cycles of high and low availability in many wild primate habitats [19,35]. ‘Thrifty genotype' models might therefore predict that frugivorous species, with their evolutionary history of ‘feast and famine' food availability, should be more susceptible to obesity in captivity than are folivorous species.

Speakman has criticized the ‘thrifty genotype' hypothesis, noting that decades of subsequent research have failed to identify plausible candidate genes [36,37]. Instead, he has argued for a ‘drifty genotype' hypothesis, that the hominin lineage experienced reduced predation risk as they became ecologically dominant, resulting in a relaxation of selection pressures limiting maximum body weight and an increased prevalence of obesity [36,37]. The ‘drifty gene' hypothesis would predict humans are uniquely vulnerable to obesity, as predation remains a salient factor for other wild primate species [38]. Further, terrestrial species of non-human primates, which are more susceptible to predation [39], would be less prone to obesity than species that are arboreal. Alternatively, arboreal species could evolve to be less prone to weight gain owing to the energy cost of carrying extra body mass up and down trees [40]. The ‘drifty gene' hypothesis would also predict that human body mass is more variable than in other species, and the distribution of body mass more right-skewed, because the mechanisms maintaining weight within a narrow range and preventing extreme weight gain have weakened.

Patterns of obesity among non-human primates offer an important point of reference for these competing explanations for the human obesity pandemic. Obesity, defined as an unhealthy accumulation of body fat, is typically identified in humans as an elevated body mass index (BMI), the ratio of (body mass)/(height²) where mass is in kg and height is in metres. Calculating BMI for other species is problematic owing to differences in body size and proportions. However, studies examining skeletal dimensions in non-human primates have reported negligible differences in the linear dimensions of the long bones [41,42]. In analyses of chimpanzees (Pan troglodytes), cotton topped tamarins (Saguinus oedipus) and Rhesus macaques (Macaca mulatta), humerus, femur and tibia lengths were either statistically indistinguishable (p > 0.05) between captive and wild populations, or were marginally (less than 5%) longer in wild populations [41,42]. The similarity in long bone lengths suggests that body lengths, analogous to ‘height’ for humans, are similar in captive and wild populations of non-human primates. If so, then it follows that the differences in body weight between captive and wild populations will correspond directly to differences in BMI, because the denominator for BMI (height or length) is similar for both groups. If a captive cohort of non-human primates has a body mass that is 50% greater than their wild counterparts, then the captive population will also have a 50% greater BMI. If we assume that populations in the wild represent the normal body weight for a given species, then relative body mass, the ratio of (captive/wild), can be used as an index of stature-adjusted weight in captive populations, analogous to BMI. As with BMI, relative body mass can be used to investigate patterns of overweight and obesity, with the important caveat that the relationship between adiposity and relative body mass is unknown.

Here, I compare body masses of captive primates from across the order, using published and publicly available measurements from zoos and research facilities. I test whether human body mass distributions differ from other primate species, and whether sex, evolved habitat preferences (terrestrial versus arboreal), or diet predicts vulnerability to obesity in captivity for non-human primates. I discuss results from these analyses and their limitations, as well as the implications for current models of human obesity.

2. Methods

(a) . Sample

I compiled published and publicly available measurements of adult body weight from both wild and captive populations (electronic supplementary material, table S1). Sampling was inherently opportunistic as it relied on data available from previous studies, and the dataset was primarily limited by the availability of body mass measures for captive species. Most measures for captive populations were calculated using data obtained from the Primate Ageing Database (https://primatedatabase.org/), an initiative of the United States National Institute on Ageing [43]. Data for wild populations were taken largely from Smith & Jungers [44] compendium, supplemented from other sources as indicated in the electronic supplementary material, table S1 and discussed below.

To calculate average adult body masses for species in the Primate Ageing Database, I calculated female and male mean masses at each age for all individuals with available data (figure 1). Most individuals were measured at several ages. Using the plot of mean body mass against age, I determined the period of prime age adulthood for each species, beginning at the end of adolescent growth and ending before age-related decline in body mass was evident. For example, for chimpanzees (Pan troglodytes) the prime adult age range was 15–35 years (figure 1). I then calculated mean body mass for every individual with at least one measurement in the prime adult age range. These mean values for each individual were then pooled to calculate female and male means and standard deviations for each species.

Figure 1.

Body mass for male and female chimpanzees in the Primate Ageing Database, binned by year of age (mean ± s.d.). Based on this plot, the age range used to calculate adult body mass was 15–30 years. Sample sizes at each age range from 22 to 74 for males and 41 to 104 for females.

Pregnancy status is not reported for the body-weight measures of the captive sample and generally unknown or unreported for the wild samples. Breeding is tightly controlled and relatively rare in United States (US) zoo populations, but it is possible that the body weights of a small number of females in both captive and wild populations were elevated owing to pregnancy. Similarly, physical activity data and population-specific diet data are not available for the large majority of populations in this analysis. I assume that captive primates are less active than their wild counterparts, and that diets in captive populations are consistent with US zoo guidelines [16] and that diets for wild populations are consistent with species descriptions in other populations.

Body masses are available for multiple wild chimpanzee populations [45]. For comparison with captive chimpanzees, I calculated mean male and female body masses using mean values from Uganda (Kibale National Park), Democratic Republic of Congo and Tanzania (Mahale Mountains National Park). Body masses from Gombe National Park, Tanzania are shown for comparison but not included in the calculation of mean wild population mass, because adults in that population are known to be smaller than typical for the species [45] and their inclusion would bias the relative body mass ratio for captive chimpanzees upwards.

For three monkey species, rhesus macaques (M. mulatta), Tibetan macaques (Macaca thibetana) and yellow baboons (Papio cynocephalus), measures of body mass are available in free-ranging populations provisioned by humans. Rhesus macaques in the Cayo Santiago research station in Puerto Rico are provisioned with 0.23 kg of primate chow per day [46]. The Tibetan macaques studied by Zhao in the 1990s near Mount Emei, China were seasonally provisioned by tourists [47]. One population of yellow baboons studied by Altmann and colleagues in the 1990s foraged regularly on human food from a rubbish dump [48]. These provisioned populations were compared to captive and unprovisioned wild populations.

Establishing human reference values for body mass is complicated by the breadth of body size diversity among human populations. While human populations vary in many different ways, the ecological distinction that is relevant for the current analysis is whether individuals produce their own food by hand, either through farming or foraging, or whether they are able to acquire industrially produced food from a market. This distinction mirrors that of non-human primates in the wild, which must find and acquire their food, versus those in captivity for whom food is provided and is often highly processed (e.g. chow). For data from human populations that produce their own food, I drew from published body mass and height data from Walker et al. [49], which compiled mean male and female cohort data for n = 22 subsistence populations. Most of these populations are short-statured compared to industrialized populations, which could bias relative body mass estimates upwards. Therefore, I limited analysis to the n = 9 populations with mean male height of 160 cm or greater, as reported by Walker et al. [49]. This subset of n = 9 populations includes groups from South America, Africa and Australia. I compared these subsistence populations to body masses from a recent study of n = 2139 men and n = 2272 women aged 18–60 in the 2003 and 2005 cycles of the US National Health and Nutrition Examination Survey (NHANES) [50]. One male with a body mass of 371 kg was considered an outlier (greater than 14 s.d. above the mean and 150 kg heavier than the next largest male) and removed from the sample. The US NHANES sample was chosen as the representative industrialized population because it is large, nationally representative, includes men and women from different races and ethnicities, and because the obesity crisis in the US is well-documented and severe. Human populations are reported in the electronic supplementary material, table S2.

For each non-human primate species, I divided mean body mass for captive males and females by the mean body mass for wild males and females, giving a ‘relative body mass' ratio, after Mellor et al. [12]. For humans, I calculated relative body mass values for men and women using each of the n = 9 subsistence populations as (US/subsistence). I calculated the coefficient of variation of body mass (standard deviation/mean) for captive primates and the US cohorts as a measure of body mass variability. I also calculated skewness of the body mass distribution for each captive non-human population with cohort sample sizes of 20 or greater, and for the US men's and women's cohorts.

I investigated the effects of sex, sexual dimorphism and ecological variables among non-human primates using diet and habitat data from the literature. Diet categories (frugivore, folivore, frugivore/folivore, omnivore) for all species were taken from DeCasien and colleagues [51], as was the percentage of fruit in the diet of wild populations when available (n = 26 species). For species without percentage dietary fruit reported by DeCasien and colleagues, I used values from the EltonTraits 1.0 database [52]. Each species was either classified as ‘arboreal' or ‘terrestrial' using ecological descriptions in the Primate Ageing Database to test whether terrestrial species, which are presumably more vulnerable to predation, are less likely to develop obesity in captivity. Humans were not included in these analyses, as the objective for these analyses was to determine patterns among non-human primates that could inform interpretations of human values.

(b) . Analyses

Analyses were conducted in R version 4.2.2 [53]. Skewness was calculated using the moments [54] package. Interspecific analyses were conducted using phylogenetic general least squares (PGLS) using the caper [55] package. Phylogenetic relationships were taken from the 10 K Trees project [56] (figure 2). To determine whether certain clades within the Primate order have become particularly vulnerable to weight gain in captivity, I used the phytools [57] package to calculate Pagel's lambda, a measure of phylogenetic signal, for female and male relative body mass ratios.

Figure 2.

Phylogeny used for this study. Asterisks (**) indicate humans (Homo sapiens) were not included in PGLS analyses.

I tested associations between relative body mass and body mass coefficient of variation with diet (both diet category and percentage of fruit) and habitat (terrestrial versus arboreal). Preliminary analyses revealed that diet (percentage of fruit) and habitat (terrestrial versus arboreal) covaried (t₃₈ = 2.27, adj. r² = 0.10, p = 0.03, λ < 0.01). Therefore, for ecological analyses I included both diet and habitat variables in the model to determine their independent effects. To assess sex effects, I tested whether female and male relative body mass ratios were correlated among species, and whether these ratios differed consistently between sexes. Finally, because body size dimorphism is related to both reproductive strategies and sex differences in growth, I explored whether body size dimorphism (mean wild male body mass/mean wild female body mass) was associated with the ratio of relative body mass (female/male) in captivity.

For both ecological and sex effects analyses, the normality of residuals was assessed using the shapiro.test function in R and visually examined using qq plots. To satisfy the assumption of normality of residuals, relative body mass and coefficient of variation values were ln-transformed for ecological analyses of diet and habitat. For all other analyses, the assumption of normality of model residuals was met (p > 0.05 Shapiro test) and transformation of variables was unnecessary.

3. Results

(a) . Human and non-human primates

The US men's and women's cohorts were larger than those from subsistence populations. US men and women were approximately 50% heavier than their counterparts in subsistence populations, with mean body mass ratios, calculated as (US)/(subsistence population) for each of the n = 9 subsistence populations, of 1.58 ± 0.20 (range: 1.17–1.82) for female cohorts and 1.44 ± 0.19 (range: 1.13–1.70) for male cohorts. These values fell within the distribution for all captive non-human primates for relative body mass (females: 1.28 ± 0.30, n = 40, range 0.67–1.78; males: 1.24 ± 0.28, range 0.70–1.97, n = 40; electronic supplementary material, table S2). The US cohorts were approximately 5% taller, with mean height ratios of 1.05 ± 0.04 (range: 0.98–1.09) for female cohorts and 1.05 ± 0.03 (range: 1.00–1.09) for male cohorts (electronic supplementary material, table S2). The ratio of US/subsistence population BMI was 1.43 ± 0.20 (range: 0.10–1.69) for female cohorts and 1.29 ± 0.18 (range: 1.02–1.53) for male cohorts (electronic supplementary material, table S2). Relative body mass (US/subsistence) was strongly correlated with the relative BMI (ordinary least squares, r² = 0.80, p < 0.01, n = 18 cohorts; electronic supplementary material, table S2).

The coefficients of variation for US women (0.26) and men (0.23) fell towards the upper end of the range of coefficients of variation in non-human primate cohorts (females: 0.15 ± 0.05, range 0.04–0.28, n = 39, males: 0.13 ± 0.04, range 0.07–0.26, n = 39) (table 1). Both male and female cohorts in the US human sample were positively skewed, indicating a longer right tail (women: 1.13, men: 1.22). Those values fell within the range of skewness for captive non-human primate cohorts (females: 0.67 ± 0.71, range: −0.97 to 1.97, n = 18; males: 0.46 ± 0.59, range: −0.56–1.46, n = 18). Among the hominoid clade, humans were most similar to captive chimpanzees and orangutans in relative body mass and coefficient of variation (table 1).

Table 1.

Key species characteristics. (Diet: O, omnivore; L, folivore; R, frugivore; R/L, frugivore/folivore. Hab.: arboreal (A) or terrestrial (T) habitat. RBM, relative body mass ratio (captive/wild). Sex dimorph: (male/female mass, non-provisioned populations). RBM ratio: female RBM/male RBM. * fruit% from [52], others from [51].)

phylogenetic group								female body mass (captivity)						male body mass (captivity)
			common	species	diet	fruit%	hab.	mean	s.d.	n	CV	skew	RBM	mean	s.d.	n	CV	skew	RBM	sex dimorph	RBM ratio
Strepsirrhine	Loridae		slender loris	Loris tardigradus	O	15	A	0.18	0.02	12	0.08		0.67	0.19	0.02	16	0.09		0.70	0.98	0.95
			slow loris	Nycticebus coucang	O	60	A	1.21	0.19	18	0.16	0.31	1.78	1.21	0.18	20	0.15	0.15	1.93	0.92	0.93
			lesser bushbaby	Galago moholi	O	0*	A	0.17	0.03	20	0.17	0.14	0.99	0.19	0.02	25	0.12	0.37	1.00	1.08	0.99
	Lemuridae		aye aye	Daubentonia madagascar.	O	0	A	2.73	0.19	16	0.07		1.10	2.61	0.18	18	0.07		1.00	1.05	1.10
			fat tailed dwarf lemur	Cheirogaleus medius	O	30*	A	0.24	0.04	38	0.18	1.67	1.41	0.24	0.04	47	0.17	0.62	1.30	1.09	1.09
			grey mouse lemur	Microcebus murinus	O	51	A	0.09	0.02	42	0.25	1.57	1.39	0.09	0.02	49	0.23	1.46	1.45	0.94	0.96
			Coquerel's sifaka	Propithecus coquereli	L	30*	A	4.19	0.43	39	0.10	0.38	0.98	3.87	0.29	32	0.08	0.66	1.05	0.86	0.94
			crowned lemur	Eulemur coronatus	R/L	30*	A	1.66	0.13	16	0.08	0.29	1.53	1.71	0.21	21	0.12	0.33	1.34	1.19	1.15
			blue eyed black lemur	Eulemur flavifrons	R/L	30*	A	2.66	0.24	18	0.09	1.23	1.51	2.49	0.32	30	0.13	−0.42	1.32	1.07	1.15
			brown lemur	Eulemur fulvus	R/L	46	A	2.61	0.47	10	0.18		1.46	2.69	0.37	14	0.14		1.42	1.06	1.03
			mongoose lemur	Eulemur mongoz	R/L	18	A	1.64	0.23	25	0.14	1.97	1.05	1.57	0.14	31	0.09	1.34	1.11	0.90	0.95
			eastern bamboo lemur	Hapalemur griseus	L	0*	A	1.04	0.17	19	0.16		1.56	1.07	0.15	14	0.14		1.43	1.12	1.08
			ring tailed lemur	Lemur catta	R	54	T	2.74	0.52	69	0.19	0.94	1.25	2.88	0.46	75	0.16	0.89	1.22	1.08	1.03
			black & white ruffed lemur	Varecia variegata	R	73.9	A	3.81	0.48	39	0.13	0.81	1.08	3.46	0.34	57	0.10	−0.34	0.95	1.03	1.13
Haplorhine	Platyrrhine	Atelidae	black howler monkeys	Alouatta caraya	L	25	A	6.78	0.84	8	0.12		1.57	11.72	1.43	5	0.12		1.83	1.48	0.86
			black handed spider monkey	Ateles geoffroyi	L	68.7	A	7.30	0.95	12	0.13		0.94	8.32	0.75	4	0.09		1.14	0.94	0.82
			brown howler monkey	Alouatta guariba	L	40*	A	4.30	0.76	13	0.18		0.99	6.70	0.48	13	0.07		1.00	1.55	0.99
			colombian spider monkey	Ateles fusciceps	R	60*	A	8.17	1.42	7	0.17		0.89	9.03	1.74	5	0.19		1.02	0.97	0.88
		Cebidae	squirrel monkey	Saimiri sciureus	O	28	A	0.68	0.15	17	0.21		1.03	1.05	0.15	17	0.14		1.34	1.18	0.77
			white fronted capuchin	Cebus albifrons	O	24.6	A	2.27	0.09	3	0.04		0.99	2.55	0.30	3	0.12		0.80	1.39	1.24
			tufted capuchin	Sapajus apella	O	36	A	3.33	0.39	4	0.12		1.32	4.83	0.38	4	0.08		1.32	1.45	1.00
			grey-bellied night monkey	Aotus lemurinus	R	20*	A	0.89		101			1.02	0.91		127			0.99	1.05	1.03
			common marmoset	Callithrix jacchus	R	22	A	0.39	0.05	281	0.14	0.08	1.02	0.38	0.04	329	0.12	0.03	1.06	0.95	0.97
			pygmy marmoset	Cebuella pygmaea	O	0	A	0.13	0.02	6	0.13		1.06	0.13	0.01	8	0.08		1.20	0.90	0.88
			golden lion tamarin	Leontopith. rosalia	O	84	A	0.66	0.08	22	0.12	−0.37	1.11	0.68	0.06	22	0.09	0.42	1.09	1.04	1.02
			cotton topped tamarin	Saguinis oedipus	O	30*	A	0.56	0.05	3	0.08		1.38	0.54	0.09	7	0.16		1.29	1.03	1.07
	Catarrhine	Colobinae	eastern black and white colobus	Colobus guereza	L	14	A	8.64	0.96	3	0.11		0.94	11.70	0.89	4	0.08		0.87	1.47	1.08
			francois langur	Trachypithecus francoisi	L	30*	A	6.43	0.89	10	0.14	−0.97	0.88	7.91	0.80	24	0.10	−0.11	1.03	1.05	0.85
		Cercopithecinae	blue monkey	Cercopithecus mitis	R	54.5	A	6.39	1.81	7	0.28		1.63	11.52	1.39	8	0.12		1.97	1.49	0.83
			Debrazza's monkey	Cercopithecus neglectus	R	77	A	7.37	0.89	6	0.12		1.78	10.66	1.00	7	0.09		1.45	1.78	1.23
			vervet	Chlorocebus aethiops	R	40*	T	5.47	0.85	356	0.15	0.71	1.66	7.62	1.25	113	0.16	1.23	1.44	1.61	1.15
			mandrill	Mandrillus sphinx	R	92	T	16.65	3.64	10	0.22		1.29	36.87	9.68	8	0.26		1.17	2.45	1.11
			yellow baboon	Papio cynocephalus	R	62	T	19.30	4.30		0.22		1.62	31.10	5.50		0.18		1.21	2.17	1.35
			rhesus macaque	Macaca mulatta	R	63	T	8.21	1.48	405	0.18	0.21	1.53	11.92	2.19	254	0.18	0.67	1.55	1.44	0.99
			chinese stump tailed macaque	Macaca thibetana	R	30*	T	16.80	2.10	13	0.13		1.77	19.50	2.20	14	0.11		1.28	1.60	1.38
		Hominidae	lar gibbon	Hylobates lar	R	60	A	7.04	1.12	4	0.16		1.32	6.86	0.80	5	0.12		1.16	1.10	1.13
			white cheeked gibbon	Nomascus leuc.	R	80*	A	7.82	1.35	7	0.17		1.07	7.95	1.58	7	0.20		1.07	1.01	1.00
			western lowland gorilla	Gorilla gorilla	L	67	T	93.52	19.29	26	0.21	0.81	1.31	170.8	29.83	32	0.17	−0.56	1.00	2.38	1.31
			human	Homo sapiens				75.26	19.68	2272	0.26	1.13	1.52	85.85	19.47	2139	0.23	1.22	1.44	1.20	1.06
			chimpanzee	Pan troglodytes	R	63	T	59.05	12.30	180	0.21	0.66	1.67	64.49	10.51	122	0.16	0.98	1.51	1.20	1.10
			orangutans	Pongo pygmaeus	R	64	A	62.11	12.10	20	0.19	0.79	1.73	121.6	15.26	16	0.13	0.51	1.55	2.19	1.12

(b) . Phylogeny and allometry

The phylogenetic signal in both relative body mass and body mass coefficient of variation for captive non-human primate populations was low. Pagel's lambda for female relative body mass (λ = 0.42, p = 0.14) exceeded that for males (λ < 0.01, p > 0.99). Similarly, Pagel's lambda for captive body mass coefficient of variation for females (λ = 0.10, p = 0.68) exceeded that for males (λ < 0.01, p > 0.99). Allometric effects were not apparent in relative body mass. Relative body mass was not correlated with body mass in the wild in PGLS analyses in female cohorts (t ₃₈= 0.62, p = 0.54, λ = 0.32, model adj. r² = −0.02) nor in male cohorts (t ₃₈= −0.17, p = 0.86, λ < 0.01, model adj. r² = −0.03).

(c) . Ecology

As noted above (Methods), diet and habitat covaried, and so were included together in PGLS analyses of body mass ratio and coefficient of variation. These analyses were performed with ln-transformed body mass ratios and coefficients of variation to satisfy the assumption of normality in the distribution of model residuals.

When diet was considered as a categorical variable, no diet category was a significant factor for ln(female body mass ratio) (folivore: reference, frugivore: t = 0.88, p = 0.38; frugivore-folivore: t = 1.49, p = 0.14; omnivore: t = 0.37, p = 0.71), but terrestrial habitat was marginally significant and associated with higher ln(female body mass ratio) (t ₃₅= 2.14, p = 0.04, λ < 0.01, model adj. r² = 0.12). Similar results were obtained when treating diet as a continuous variable. Fruit percentage was not a significant factor for ln(female body mass ratio) (t = 1.13, p = 0.26), but terrestrial habitat was marginally significant and associated with higher ln(female body mass ratio) (t ₃₇= 2.03, p = 0.05, λ < 0.01, model adj. r² = 0.13). For male cohorts, neither diet category (folivore: reference, frugivore: t = 0.79, p = 0.43; frugivore-folivore: t = 0.95, p = 0.35; omnivore: t = 0.29, p = 0.77) nor terrestrial habitat (t ₃₅= 0.57, p = 0.57, λ < 0.01, model adj. r² = −0.05) were associated with ln(male body mass ratio). Similarly, neither fruit percentage (t = 1.21, p = 0.23) nor terrestrial habitat (t ₃₇= 0.40, p = 0.69, λ < 0.01, model adj. r² < 0.01) were associated with ln(male body mass ratio).

In analyses of body mass variability, ln(female coefficient of variation) was not associated with diet category (folivore: reference, frugivore: t = 0.72, p = 0.47; frugivore-folivore: t = −0.79, p = 0.43; omnivore: t = −0.77, p = 0.44) nor with terrestrial habitat (t ₃₄= 1.07, p = 0.29, λ < 0.01, model adj. r² = 0.11). In analyses using fruit percentage as the dietary variable, fruit percentage was positively, though marginally, associated with ln(female coefficient of variation) (t = 2.21, p = 0.03), but terrestrial habitat was not a significant factor (t ₃₆= 1.48, p = 0.15, λ < 0.01, model adj. r² = 0.18). For males, ln(male coefficient of variation) was not associated with diet category (folivore: reference, frugivore: t = 1.47, p = 0.15; frugivore-folivore: t = 1.07, p = 0.29; omnivore: t = 1.24, p = 0.22) but was positively associated with terrestrial habitat (t ₃₄= 2.75, p = 0.01, λ < 0.01, model adj. r² = 0.21). In analyses using fruit percentage as the dietary variable, fruit percentage was marginally, positively associated with ln(male coefficient of variation) (t = 2.06, p = 0.05) and terrestrial habitat was also positively associated as well (t ₃₆= 2.65, p = 0.01, λ < 0.01, model adj. r² = 0.28).

(d) . Sex and sexual dimorphism

A paired t-test revealed no difference between male and female relative body mass (t ₄₀= 1.51, p = 0.14). Female and male relative body mass ratios were correlated in PGLS analyses (adj. r² = 0.66, t ₃₉= 8.81, λ = 0.32, p < 0.001), as were male and female coefficients of variation (adj. r² = 0.36, t ₃₈= 4.83, λ < 0.01, p < 0.001). When fruit percentage and habitat were included as covariates in the model, female body mass ratio remained correlated with male body mass ratio (t = 9.05, p < 0.001), fruit percentage was not a significant factor (t = 0.20, p = 0.84), and species with terrestrial habitats had higher female body mass ratio (t ₃₆= 3.31, λ < 0.01, model adj. r² = 0.73). When fruit percentage and habitat were included as covariates in a multivariate model for coefficient of variation, neither fruit percentage (t = 1.27, p = 0.21) nor habitat (t = 0.43, p = 0.67) were significant factors (model adj. r² = 0.31, λ < 0.01).

Sexual dimorphism in body mass for wild populations was associated with the ratio of female/male relative body mass in captive populations. In PGLS analyses, females of species with greater body mass dimorphism in the wild had greater relative body mass ratios in captivity (t ₃₈= 3.96, p < 0.001, λ = 0.06, adj. r² = 0.27). This association remained significant (t ₃₆= 2.65, p = 0.01, λ = 0.03, adj. r² = 0.30) when fruit percentage (t = −0.74, p = 0.47) and habitat (t = 168, p = 0.10) were add to the model as covariates.

(e) . Provisioning

Comparing body mass means for species with multiple populations, it is evident that provisioned non-human primate cohorts, either in captivity or free-ranging, have greater body mass than cohorts who work to acquire their food (figure 4). Body mass differences between chimpanzee populations in captivity and those in the wild were similar in magnitude to differences observed between the USA and subsistence populations (figure 4). Baboons in the wild with access to a rubbish dump weighed more than those without, and were similar in body mass to baboons in US research facilities (figure 4). Wild stump tailed macaques provisioned by tourists were heavier than those which were not (figure 4). Free-ranging macaques at the Cayo Santiago research colony were heavier than those in the wild and similar in body mass to those in US research facilities (figure 4).

Figure 4.

Body mass (mean + s.d.) for female and male cohorts of humans, chimpanzees, yellow baboons, stump tailed macaques and rhesus macaques. Provisioned populations (including humans in the industrialized US) are in black. See data in the electronic supplementary material, tables S1 and S2.

4. Discussion

Provisioned primates generally weigh more than those that must work to acquire food from their environment. The magnitude of this effect varies, but species with relative body masses similar to that observed in comparisons of industrialized and subsistence human populations are found across the Primate order (figure 3 and table 1) and across different ecologies (figure 3). Relative body mass for people in the US, an industrialized population, measured against men and women from nine subsistence populations, is similar to our close evolutionary relatives (chimpanzees, orangutans) as well as more distantly related monkeys, lemurs and lorises. Variability in body weight, as assessed by body mass coefficient of variation, is also similar. Humans do not appear to be unique in our propensity to develop obesity.

Figure 3.

Relative body mass (box indicates 25th and 75th percentiles, inner line indicates median, whiskers indicate range, circles indicate outliers) for female (white) and male (grey) cohorts across (a) diet groups, (b) habitat types and (c) phylogenetic groups.

Relative body mass appears to provide a useful, if approximate, measure of stature-adjusted weight for non-human primates, analogous to BMI for humans. Among human cohorts in this study, relative body mass and relative BMI were strongly correlated (r² = 0.80) because variation in stature was relatively low. Previous studies in chimpanzees, Rhesus macaques and cotton topped tamarins have shown that limb bone lengths, an index of stature, are similar between captive and wild primate populations. Relative body masses for these species are similar to those of US cohorts, suggesting similar prevalence of overweight and obesity. Of the 41 non-human primate species in this study, 13 female and six male cohorts had relative body masses greater than 1.50, suggesting overweight and obesity prevalence in the human range. As with measures of BMI in humans, we do not know whether the elevated body mass in these non-human primate cohorts corresponds closely to increased adiposity.

Human cohorts and most others with sufficient data to estimate skewness had longer right tails in their body mass distributions, indicating that positive deviations from mean body mass (i.e. overweight and obese individuals) were both more common and more extreme than negative deviations (i.e. underweight). The skewness of non-human primate cohorts in captivity is probably constrained by carer efforts to prevent or reduce obesity and overweight. Extreme body weights are less likely to occur under veterinary care, because carers would intervene. Skewness values are also sensitive to outliers. For example, if the largest chimpanzee body mass on record (a male named Kermit, approx. 204 kg [58]) is included, the skewness of the male chimpanzee cohort increases from 0.98 to 2.08. Similarly, including the largest captive gorilla on record (a male named Samson, 296 kg [59]) increases the skewness for male gorillas from −0.56 to 0.66. Free-living humans are not under the supervision or dietary control of carers, and are therefore more able to obtain extreme weights. The comparatively high skewness of the US cohorts can therefore be attributed to the lack of intervention to prevent or reduce extreme obesity in a relatively small number of adults.

Results of this study challenge evolutionary hypotheses for the global obesity pandemic that propose humans are particularly vulnerable to obesity owing to recent changes in predation risk [36,37] or in the severity or regularity of famines [33]. There is no evidence for human exceptionalism in the present analysis. ‘Thrifty gene’ hypotheses also predict that species with more variability in their evolved diet should be more vulnerable to obesity when food availability is chronically high. However, we find little evidence that frugivorous non-human primates are more susceptible to obesity in captivity, or that folivores are protected. ‘Drifty gene' hypotheses, which propose that predation risks impose selection pressure to limit body mass and maintain mobility [37], would suggest that terrestrial primates, which are exposed to greater predation [39], would be less prone to obesity. We see no evidence for that among non-human primates.

Nor do we see evidence that Palaeolithic dietary adaptations make humans uniquely vulnerable to overconsumption today. Ethnographic and archaeological evidence indicates that past human diets were variable over time and geography, and were not particularly low in carbohydrates [60]. Nonetheless, the evolution of hunting and gathering increased the proportion of meat in the human diet relative to the diets of other living apes [60], and some have argued that humans are adapted to low-carbohydrate, meat-heavy diets that make us vulnerable to cardiometabolic disease caused by more carbohydrate-rich diets today [28,61,62]. Results here challenge that hypothesis. Humans are not uniquely susceptible to overconsumption, and we see no evidence for diet effects on relative body mass among non-human primates. The capacity for human-like obesity appears to be shared widely across primates, a largely frugivorous and folivorous order roughly 65 million years old [63].

The similarity in body mass ratios and body-weight variability among human and non-human primates suggests common factors may contribute to obesity across different species and settings. One point of similarity may be diet. Chow-based diets in captivity tend to have greater energy density (MJ g⁻¹) [17] and contain more fat and processed foods than wild diets. Diets in industrialized human populations also tend to be higher in energy density and fat, and contain less ultra-processed food, than the diets of subsistence populations [60,64]. A better understanding of the mechanisms by which these aspects of diet impact consumption could clarify their role in the aetiology of obesity for both humans and non-human primates.

Anxiety, boredom, depression, or other psychosocial stresses might also be greater in both captive primate and in industrialized human populations, and could contribute to the development of obesity across species [30–32]. Studies are needed to test whether these potential stressors are in fact greater in captive and industrialized contexts. It is notable that provisioned populations of baboons [48] and stump tailed macaques [47,65] in the wild, and of free-ranging rhesus macaques in the Cayo Santiago research colony all achieve relative body masses similar to those in research facilities and other primate populations in zoos (figure 4). These comparisons suggest that captive settings in the US are not uniquely obesogenic environments for primates.

Relative body mass and body-weight coefficient of variation were not uniform across species in this analysis. The underlying causes of this interspecific variability in relative body mass are unclear. Differences in animal care across facilities could be one factor. Many different zoos and other facilities are represented in this dataset, and nutritional care is not uniform. Variation in staffing, resources, training, and approach could all lead to variability in weight management.

The evidence for ecological factors underlying the variability in body weight among captive non-human primate populations was weak and inconsistent. The only association with evolved diet was a statistically marginal, positive correlation between the percentage of fruit in the diet and coefficient of variation in body mass among female cohorts. Terrestrial (versus arboreal) habitats were marginally associated with greater relative body mass in female cohorts and greater body mass coefficient of variation in male cohorts. The tendency for female cohorts of terrestrial species to develop greater relative body masses, and for both males and females of these species to have greater inter-individual variability, could be related to locomotor performance. Increased body mass might compromise agility or leaping performance in the tree canopy, and has a greater impact on the energy cost of climbing than on the energy cost of walking [40,66]. Thus, it is possible that selection on locomotor performance has resulted in constraints on body mass among arboreal species that are relaxed for terrestrial species, as argued by Heldstab et al. [40], but the tenuous nature of the results here indicate additional testing is needed to determine if these habitat effects are robust.

Male and female cohorts did not differ consistently in relative body mass, and instead were strongly correlated, indicating that the physiological mechanisms underlying overconsumption in captivity are largely shared between sexes. Nonetheless, sexual size dimorphism was clearly associated with sex differences in relative body mass across species (figure 5). The smaller sex gained proportionally more weight in captivity, and the magnitude of the effect was related to the degree of sexual dimorphism. A possible explanation for this phenomenon is that the larger sex monopolizes less of the food in captivity, such that the smaller sex gains proportionally more weight. Notably, the similarity in men's and women's relative body mass in humans is consistent with our species' low degree of sexual dimorphism.

Figure 5.

The ratio of female relative body mass/male relative body mass plotted against sexual dimorphism (mean male mass/mean female mass, wild populations) for non-human primates.

The analyses here are limited by sample size. There are over 300 recognized primate species [56], and only 41 are included here. Greater and more public data sharing across zoos, sanctuaries and other facilities would increase the breadth of species available for these and other studies. Sample sizes within many of the primate cohorts, particularly for those in the wild, are often small as well. Additional measures of wild body mass would improve estimates of relative body mass.

Analyses here are also limited by a lack of body composition data. Because captivity does not appear to increase body length, measures of relative body mass for captive primates provide a measure of body size that is analogous to BMI. However, as with human studies that rely on BMI, we cannot assess adiposity for the non-human primates in this sample. Studies of captive chimpanzees [8], wild baboons [48] and humans [67] suggest that body mass is positively related to adiposity among primates, but more data is needed. Clearly, provisioned primates gain substantial fat free mass as well as fat mass in captivity. For example, fat free masses for male (approx. 55 kg) and female (approx. 43 kg) chimpanzee cohorts in zoos and sanctuaries [17] exceeds total body mass for cohorts in the wild ([45]; figure 4). Similarly, fat free mass of men and women in industrialized societies exceeds that of subsistence populations by approximately 30% [17]. Primate species differ in their degree of adiposity in captivity [10]. Measures of body composition would enable researchers to examine the degree to which weight gain is channelled towards fat free mass or fat mass across species, and to better identify individuals with unhealthy levels of adiposity.

Patterns of body mass across captive primate cohorts provide additional perspective on the global obesity crisis facing humans. Our species is not unique in its propensity to put on weight when provisioned. Results here hint at the possibility that our evolved terrestriality may contribute to the human tendency towards weight gain, as it appears to in other species, but additional work is needed. Zoos and sanctuaries may be important and fruitful settings for future work, allowing researchers to test the effects of dietary composition (e.g. macronutrient proportions, energy density) and psychosocial factors on eating behaviours and cardiometabolic health. This work would not only be for human benefit but could also improve the health of study populations in captivity. The broad interspecific comparisons here highlight the use of drawing on a wide range of species with diverse evolutionary histories to better understand ourselves.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

All data are available in table 1 and the electronic supplementary material, table S1 and S2 [68].

Declaration of AI use

I have not used AI-assisted technologies in creating this article.

Authors' contributions

H.P.: conceptualization, data curation, formal analysis, investigation, methodology.

Conflict of interest declaration

I declare I have no competing interests.

Funding

I received no funding for this study.