Modelling the maximal active consumption rate and its plasticity in humans—perspectives from hot dog eating competitions

Abstract

Gut capacity and plasticity have been examined across multiple species, but are not typically explored in the context of extreme human performance. Here, I estimate the theoretical maximal active consumption rate (ACR) in humans, using 39 years of historical data from the annual Nathan's Famous Hot Dog Eating Contest. Through nonlinear modelling and generalized extreme value analysis, I show that humans are theoretically capable of achieving an ACR of approximately 832 g min⁻¹ fresh matter over 10 min duration. Modelling individual performances across 5 years reveals that maximal ACR significantly increases over time in ‘elite’ competitive eaters, likely owing to training effects. Extreme digestive plasticity suggests that eating competition records are quite biologically impressive, especially in the context of carnivorous species and other human athletic competitions.

1. Introduction

The maximal limits of human performance are a perpetually interesting, interdisciplinary topic of scientific discourse. Digestive system capacity is not typically explored as a measure of human performance, but the ability to consume and digest large quantities of food is advantageous if resource availability is unpredictable (i.e. predator–prey interactions, carrion acquisition) [1]. Additionally, the ability to rapidly ingest large quantities of food (i.e. time-minimizing eating behaviours) may be advantageous under certain conditions (e.g. conspecific or heterospecific competition, food scarcity). This can be quantified as the active consumption rate (ACR), defined as the mass of food consumed in a given active feeding time period [2].

While a large ACR in a short timespan could be advantageous in certain environments, excess gut capacity is metabolically expensive to maintain. In humans and some other primates, gut size is relatively reduced, which presumably allowed energy to be allocated to encephalization [3]. However, morphological adaptations could theoretically make it difficult to adapt to large fluctuations in food abundance. Some species accommodate for this through shorter term phenotypic plasticity, in that digestive morphology within an individual is altered in relation to current dietary behaviours [4,5]. Thus, a relatively small gut capacity combined with the ability for significant plasticity may be beneficial to an organism's ability to adapt to changing environments. Though ACR has been studied in other species, this metric of gut capacity has not been analysed humans.

Thus, the purpose of this study was to provide the first estimate of maximal ACR and its plasticity in humans. Here, I use historical data from Nathan's Famous Hot Dog Eating Contest (heretofore, referred to as ‘contest’) to estimate the maximal limits of ACR in humans and compare it to known values from terrestrial carnivores. This competition, held every July 4 in New York City, features participants attempting to consume as many hot dogs (with buns) as possible within a fixed time limit (generally 10 or 12 min). Though often perceived as an entertaining spectacle of gluttony, nearly four decades of data from the event provide insight into the limits of human gut capacity and its intra-individual plasticity. I use previously established mathematical models [6] to determine how many hot dogs a human can rapidly consume and demonstrate that this is owing to plasticity in gut capacity (i.e. a training effect).

2. Material and methods

(a). Data compilation and eligibility

Historical data were collected from the contest's website [7]. Attempts were made to find primary news sources (e.g. newspaper and magazine articles) to verify data from the website and competition time limits, and obtain body mass and height of the winners.

Confirmed data from years in which the competition duration was 10–12 min were included in analyses. Though it officially began in 1972, the contest duration was only 3.5 min. Data from 1973 and 1975–1977 were not available. In 1978 and 1981, the competition durations were less than 10 min. The 1979 dataset was excluded owing to inconsistencies in reported results. Thus, 39 years of data were eligible for analysis (1980, 1982–2019), accounting for 152 distinct competitors.

ACR was initially computed as number of hot dogs consumed divided by event duration.

(b). Mathematical modelling

Historical progression of athletic records often follows a sigmoidal pattern, featuring an initial slow and steady rise in performance, followed by an era of rapid improvement, and finally an apparent plateau [6]. A model to determine the theoretical maximum ACR was developed based on the two-step procedure employed by Denny [6] to determine maximal running speed across humans, dogs and horses using historical competition data. First, a best-fit nonlinear regression model is developed from historical data. Second, generalized extreme value (GEV) analysis of the residuals (model-predicted versus observed data) is used to estimate the maximal possible deviation from model-predicted values. This value is then added to the greatest model-predicted value (i.e. the asymptote) to estimate the upper limit of performance.

Data were modelled using logistic regression (equation (2.1)):

$$\text{ACR}(y)=mn+(mx-mn)\left(\frac{\exp [k(y-t)]}{1+\exp [k(y-t)]}\right).$$ 2.1

This equation is useful for modelling data that eventually reach a maximal limit over time. ACR(y) represents the contest winner's predicted ACR for a given year, mn represents the lowest ACR, mx represents the greatest ACR recorded (i.e. the contest record). k is a parameter used to adjust the shape of the plot (i.e. how quickly the ACR transitions from the early plateau to the later plateau). Y is the year at which the increase in ACR was most rapid. The model was constrained, so that mn equalled the lowest historically observed winning ACR (0.833 hot dog min⁻¹).

Data were fit to this model using nonlinear regression using SPSS v. 26.0 (IBM Corp., Armonk, NY). The least-squares criterion of fit was used to estimate mx and its 95% confidence interval. Bootstrapping was used to compute standard errors for all parameters.

As noted by Denny [6], this model estimates mx from the existing data, but does not account for extreme values. Historical data points may exceed model-predicted values. To account for extreme values, the GEV distribution was used. Residuals from the nonlinear model were computed and analysed using equation (2.2), where a is the shape parameter, b represents the location parameter and c is the scale parameter.

$$P(\text{ACR})=\text{exp}\left\{-{\left[1+a\left(\frac{\text{ACR}-b}{c}\right)\right]}^{-1/a}\right\}.$$ 2.2

Under the appropriate conditions, the absolute maximal extreme value can be determined (equation (2.3)):

$$\hspace{0.17em}\hspace{0.17em}\mathrm{AC}{\text{R}}_{\max }=b-\frac{c}{a}.$$ 2.3

GEV distribution parameters were determined using extRemes [8] and in 2 extRemes packages [9] for R v. 3.5.2 [10].

(c). Estimation of active consumption rate

Nutritional data from Nathan's Famous were used to estimate hot dog mass, and nutrient and energy composition [11]. The contest uses a 8 : 1 natural casing hot dog with bun, with a mass of 100 g [12]. The number of hot dogs was multiplied by this value to determine the absolute ACR.

ACR was also normalized to body mass (when available) and computed in relation to total energy content. The combined hot dog and bun are approximately 53% dry matter and approximately 290 kcal (approximate energy content, 55% fat; 31% carbohydrate; 14% protein). Details of food composition and energy content for hot dogs are provided in the electronic supplementary material, methods.

Hot dog composition and size have remained unchanged in the company's 103 years [12]. This consistency allows for valid ACR comparisons between individuals within one competition, and within and between individuals across years of competition.

(d). Intra-individual changes in active consumption rate

To determine if ACR increases over time, individuals who competed in at least five contests were identified. ACRs were computed for each individual's first five competitions. To minimize the effect of competition duration (i.e. 10 versus 12 min), only individuals who had at least four competitions of equal duration were included in the analysis, and ACRs were only computed for these competitions. Individuals with two competitions of one duration and three of another duration were excluded.

Separate linear mixed effect models were developed for men, women and combined sexes. ACR served as the dependent variable. Participant served as a random factor. Competition number (i.e. first through fifth) served as a covariate. Models were fit using multiple repeated measures covariance matrices, and the model with the lowest Aikake's information criterion was chosen as the best fit [13]. Statistical significance was set as p < 0.05.

3. Results

Parameter estimates for nonlinear regression are presented in table 1, with mx computed to be 7.34 hot dogs min⁻¹. Parameter estimates for GEV are presented in the electronic supplementary material (electronic supplementary material, table S1). Using these parameters to solve for ACR_max yields a maximum extreme residual of 0.983 hot dogs min⁻¹. When added to the mx of the nonlinear model, this yields an ACR limit of 8.32 hot dogs min⁻¹, which converts to 832 g min⁻¹. Over 10 min, this is an energy intake of 24 000 kcal. Body mass normalized and energy-referenced ACRs of selected winners are presented in the electronic supplementary material.

Table 1.

Parameter estimates of nonlinear modelling of ACR. s.e., standard error; mx, model-predicted maximal ACR; mn, constrained model minimum ACR; k, model shape parameter (speed that mn transitions to mx); t, year of most rapid increase in ACR.

	estimate	s.e.	95% confidence interval
			lower	upper
mx	7.34	0.32	6.81	7.86
mn	0.83
k	0.222	0.03	0.162	0.282
t	2003	1	2002	2005

Data points from contest winners and the nonlinear regression curve are plotted in figure 1.

Figure 1.

Active consumption rate (ACR) of the winner of the Nathan's Famous Hot Dog Eating Contest by year. Circles represent 10 min competitions, squares represent 12 min competitions. The spline represents predicted ACRs based on the nonlinear model.

Seventeen males and 7 females met the criteria for linear mixed effects models. There was a significant positive effect of competition number for men and combined sexes, but not for women (table 2).

Table 2.

Parameter estimates of linear mixed effects models for the effect of competition number on ACR. s.e., standard error.

sex	parameter	estimate	s.e.	95% confidence interval
				lower	upper	p-value
men (n = 17)	intercept	2.53	0.17	2.17	2.88	<0.001
	competition	0.14	0.05	0.04	0.24	0.007
women (n = 7)	intercept	1.99	0.39	1.15	2.84	<0.001
	competition	0.08	0.07	−0.07	0.22	0.292
combined (n = 24)	intercept	2.40	0.15	2.10	2.71	<0.001
	competition	0.11	0.04	0.03	0.19	0.008

4. Discussion

This paper presents the first estimates of ACR in humans. Historical data from Nathan's Famous Coney Island Hot Dog Eating Contest suggest that the upper limit of fresh matter ACR in humans is 832 g min⁻¹. Trained competitive eaters regularly reach ACRs greater than 400 g min⁻¹ with current record of 734 g min⁻¹, representing more than 21 000 kcal consumed in 10 min. If one assumes that early contest winners represent the high-end of ‘untrained’ humans, an average person may be capable of a fresh matter ACR of approximately 100 g min⁻¹ and consume approximately 3000 kcal in 10 minutes. When normalized to body mass, this translates to ACRs of approximately 0.8–1.6 g kg⁻¹ min⁻¹, corresponding to 2.4–4.6 kcal kg⁻¹ min⁻¹ for the average person and 10.5 g kg⁻¹ min⁻¹ (30.5 kcal kg⁻¹ min⁻¹) in one elite competitor (details in electronic supplementary material, results).

(a). Comparison between species

Rapidly consuming large quantities of food can be ecologically beneficial. For instance, carnivores that kill prey exceeding individual gut capacity may share meat and/or need to hunt less frequently to achieve energetic maintenance requirements [14]. Thus, the capacity to achieve a high ACR could have been advantageous in human evolution. The contest duration makes it difficult to directly relate ACRs between species, though comparison with existing data highlights the impressive eating capacity of humans. Maximal absolute ACRs in humans are similar to that observed for grizzly bears eating muscle (798 ± 246 g min⁻¹), though smaller than that for grey wolves (1119 ± 152 g min⁻¹) [2], although these were measured in feeding durations of 0.5–6.0 min (CC Wilmers, unpublished data from [2], provided by personal communication, 19 March 2019). When normalized to body mass, maximal human ACR (approx. 10 g kg⁻¹ min⁻¹, see electronic supplementary material, results) exceeds bears (2.8 g kg⁻¹ min⁻¹), but is far lower than wolves (24.8 g kg⁻¹ min⁻¹) for these short durations. Human ACR is greater than coyotes' (approx. 11 kg body mass), which achieve an ACR of approximately 3.5 to 5.8 g kg⁻¹ min⁻¹ during approximately 6–10 min bouts of muscle consumption [2] (CC Wilmers, unpublished data from [2], provided by personal communication, 19 March 2019).

(b). Comparison to athletic competitions

Models developed from 39 years of data from the eating competition (figure 1) appear to follow a similar pattern to other athletic competitions [6]. In running competitions, this pattern is attributable to a combination of increased participation, greater incentive for maximizing performance, improved training techniques and competitive strategies, and new technology. Similar factors likely influenced the contest as it transformed from a local holiday festivity into a serious international competition with qualification standards and monetary prizes.

As with major athletic competitions, elite participants engage in specialized training (i.e. rapidly consuming large volumes of food or fluid) intended to increase performance [15,16]. For instance, the contest's record holder achieved a debut performance of 267 g min⁻¹ in 2005, and incrementally increased his ACR over the next 4 years to 680 g min⁻¹, before eventually reaching 740 g min⁻¹ in 2018 (electronic supplementary material, figure S2). Other top male and female contestants have demonstrated considerable improvement in their first 5 years of competition (figure 2). Indeed, ACR increases with subsequent competitions in males and combined sexes. The lack of significant effect for females is likely owing to insufficient sample size, but could also be attributable to differences in motivation, training practices, or anatomy/physiology.

Figure 2.

Progression of ACR in the first 5 years of competition for individuals who have achieved one of the top 50 performances in contest history. While faster eating technique may account for some of these increases, plasticity of digestive capacity owing to load-induced adaptation likely plays a major role to increase ACR.

The feats of elite competitive eaters are quite biologically impressive when placed in the context of other sports competitions. Winning ACR has increased approximately 700% in less than 40 years, whereas world record performances in various mainstream sports have only progressed a median of approximately 40% since the inception of world record recognition (median year = 1922, electronic supplementary material, figure S1). Many of today's elite competitors achieve an ACR fivefold above the average ‘untrained’ individual (i.e. early contest winners). This starkly contrasts marathon competitions, where the world record holder ‘only’ maintains approximately double the speed of average marathoners [17–19] and is still not quite fivefold faster than an individual briskly walking the distance at 5 km h⁻¹.

(c). Physiology of extreme active consumption rates in humans

Human performance is typically enhanced through physiologic stimuli/stressors (i.e. exercise training) activating a complex series of molecular signalling pathways, which ultimately enhance function at cellular, tissue and system level [20]. Conversely, medical imaging suggests digestive plasticity in competitive eaters may be owing to physiologic dysfunction. An elite competitive eater experiences a decreased resting gastric emptying rate and extreme gastric dilation, severely decreased peristalsis and ablation of normal physiological signals of satiety compared to a healthy individual following 10 min of rapid hot dog consumption [15]. The large boluses of food ingested by competitive eaters remain in the digestive tract for days before excretion [15,21], and severe, self-resolving gastric distension has been reported in a competitive eater [21]. Such apparent dysfunction is consistent with clinical populations. Obese humans have a gastric capacity approximately double that of lean individuals [22] and those who engage in binge-eating behaviour have greater gastric capacity than those who do not [23]. Likewise, bulimic women engaging in binge-eating behaviour have greater gastric capacity and delayed gastric emptying compared to age- and size-matched non-binge eaters [24].

Thus, ACR ‘improvements’ across competitors arguably represent intra-individual phenotypic plasticity in response to long-term environmental (i.e. dietary) conditions [1,25,26], which are advantageous for competitive eating success but potentially pathologic or pre-dispose to other pathologies [16]. However, there are insufficient data to assess the long-term health-related outcomes of trained competitive eaters. If chronic damage, rather than adaptation, is responsible for extreme gut capacity, this may explain why the progression of eating records exceeds that of athletic competitions, as a bioenergetically costly and highly regulated load-adaptation cycle is not necessary to optimize ACR.

(d). Factors limiting active consumption rate

Extreme feeding events typically trigger signals of satiety, but an absence of such normal signals (e.g. large prey carnivores not feeding to gut capacity) may prompt deleterious behaviours and lead to metabolic dysfunction [14]. For human competitive eaters, the opposite may be true, such that extreme meals apparently seemingly impair normal satiety physiology, allowing unnaturally high ACRs. Unfortunately, there are currently insufficient historical competition data available to allow maximal ACR to be modelled from energetically and structurally heterogeneous diets. However, case comparisons may provide some insight into how these influence ACR.

Nutrient and dry matter content could have a large influence on satiety/gastrointestinal tolerance signals and influence ACR [27–29]. For instance, one competitor achieved a fresh matter ACR of approximately 408 g min⁻¹ in the 2007 contest, but only achieved one-third of that while setting the chocolate eating record (unspecified cacao percentage, approximately 98% dry matter) a year earlier (895 g in 7 min; ACR approx. 127 g min⁻¹) [30]. If dry matter ACRs are compared for each, the gap is much reduced (216 g min⁻¹ versus^. approx. 124 g min⁻¹). These differences are similar when expressed in total energy content (approx. 1200 kcal min⁻¹ versus approximately 660 kcal min⁻¹). On dry matter basis, fat content is similar between both (55% versus approx. 52%), but carbohydrate content differs considerably (31% versus approx. 43%), which could influence biochemical satiety signals.

The material properties of food, which are influenced by preparation methods, can influence chewing mechanics and energetics, and thus ease of consumption [31,32]. Thus, eating competitions involving bone-in meats, especially from wild game, may be more ecologically/evolutionarily relevant and are likely to yield lower values than those for hot dogs. For instance, the 2012–2013 winner achieved hot dog ACRs of 567–575 g min⁻¹ over 12 min. During this same era, he set records for consuming 6.25 kg of pork rib meat in 12 min (2013) and 3.46 kg of chicken wing meat in 12 min (2012) [30]. These yield fresh matter ACRs of 521 g min⁻¹ and 288 g min⁻¹, respectively. Discrepancies are likely explained by chewing limitations, since the dry matter content of all foods is roughly similar (approx. 45–60%). Likewise, biochemical satiety signals were unlikely to limit ACR, since hot dogs have the greatest carbohydrate content. Nonetheless, it remains unknown whether such differences in ACR between foods are primarily attributable to physical structure or nutrient/energy content, especially given the likelihood of physiologic dysfunction in competitive eaters.

(e). Limitations

Eating competitions do not reflect actual ecological conditions—a known limitation when studying athletes in the context of evolutionary biology [33]. These contests provide each individual with an unlimited, ready-to-consume food supply. Thus, participants can focus all of their efforts on maximizing consumption, rather than investing energy into foraging, chasing prey or competing with others for access to a dwindling supply. Further, the presence of spectators may also influence ACR [34].

The composition of hot dogs and buns has been reported to be consistent over the contest's history, but it is possible that ingredient composition could have changed (e.g. changes in cereal meal used for bun, changes in beef composition owing to agricultural practices). This could potentially influence performance trends over four decades, but seems unlikely to be the major contributor to the approximately 700% increase observed.

5. Conclusion

In summary, data from hot dog eating competitions suggest that there is stunning plasticity in gut capacity, with fresh matter ACRs of 832 g min⁻¹ theoretically possible. The rate of performance progression of competitive eaters far exceeds that from athletes in mainstream sports, but the physiological adaptations to achieve this may be indicative of training-induced digestive dysfunction.

Ethics

This article does not present research with ethical considerations.

Data accessibility

The full dataset is provided as electronic supplementary material.

Competing interests

J.M.S. has no competing interests, financial or otherwise, to declare.

Funding

J.M.S. did not receive any external funding in relation to this manuscript.