The interaction between exercise, appetite, and food intake has received widespread scientific attention within recent years given the direct relevance for energy balance and weight control (4). Emergent from this body of research is a consensus that single bouts of moderate- to high-intensity exercise transiently suppress appetite but have no influence on ad libitum energy intake (10,33). Energy homeostasis therefore seems insensitive to acute energy deficits imposed by exercise, with more prolonged or repeated perturbations necessary to induce partial compensatory responses (36,39). In association with this line of research has been a related interest in seeking to understand the mechanisms underpinning appetite control and perturbations in energy balance resulting from exercise and dietary interventions. Notably, the responses of several gut peptides to exercise (acylated ghrelin, peptide YY3–36, glucagon-like-peptide-1, and cholecystokinin) have been scrutinized as possible modulators of appetite and food intake (34). The most consistent finding from these investigations is that exercise transiently alters the circulating concentrations of these hormones in directions associated with suppressed appetite; however, circulating concentrations are typically not different from control at 30 to 60 min postexercise (10).
With a growing emphasis within biomedical science on “precision medicine” (2), recent research has sought to characterize the individual variability in appetite and energy intake responses to exercise (13,18,20,27). The primary question addressed within these studies is whether some individuals are more or less likely to compensate for energy expended during exercise by increasing postexercise energy intake. The implication of this inquiry is that exercise may be less useful for weight management in “compensators” compared with “noncompensators.” Unfortunately, to date, the studies that have examined this issue are limited by small sample sizes and the failure to appreciate the importance of internal sources of variation (technical error and biological variation) (1). Additional research is therefore needed to provide greater insight into this area of research.
For the last 15 years, our research group has conducted many experimental exercise interventions examining the effects of acute exercise on appetite, ad libitum energy intake, and appetite-regulatory hormones. Given the uniqueness of acylated ghrelin as the only circulating hormone known to stimulate appetite and promote positive energy balance (9,40), our research has maintained a central focus on the interaction between exercise, appetite, ad libitum energy intake, and acylated ghrelin. Usefully, the experimental designs (randomized crossover trials with exercise and control trials), participants (lean, young, healthy males) and exercise protocols (aerobic moderate- to high-intensity exercise) utilized within these studies have been remarkably similar. This similarity permits the aggregation of data, which provides enhanced power to investigate experimental intervention effects and to interrogate associations between key variables. Uniquely, in this context, this large data set also provides a novel opportunity to comprehensively explore the variability in appetite and ad libitum energy intake responses to exercise between individuals.
The primary aims of this study were twofold. First, using our large, pooled data set of experimental trials, we sought to characterize the immediate (during and shortly after exercise) and extended (several hours postexercise) effect of acute exercise on perceived hunger, ad libitum energy intake and circulating concentrations of acylated ghrelin. Second, with the precise consideration of the day-to-day biological and technical error inherent within outcome measurements, we sought to determine the individual variation in hunger, ad libitum energy intake and circulating acylated ghrelin responses, both during and in the hours after a single bout of exercise. To achieve this second aim, we have collected new data to determine the day-to-day variation (with no intervention) in hunger, circulating acylated ghrelin and energy intake (during ad libitum feeding) in young, healthy males. The findings reported in this manuscript provide novel insights concerning the interaction between exercise, appetite control and energy homeostasis.
METHODS
Research studies and participants
The data described in this manuscript were derived from 17 studies (all published in peer-reviewed scientific journals), which were conducted between 2004 and 2014 in the exercise physiology laboratory led by Professor David Stensel at Loughborough University, United Kingdom. All included studies received ethical approval from the institutional ethical advisory board, and written informed consent was obtained from all participants before any trial procedures commenced. Each trial included within this pooled analysis was an acute randomized crossover trial with participants having completed paired exercise (see next section) and control (resting within the laboratory) trials. The key features of each study in this pooled investigation are described in tables within the accompanying Supplementary Digital Content (1–8, https://links.lww.com/MSS/A855, https://links.lww.com/MSS/A856, https://links.lww.com/MSS/A857, https://links.lww.com/MSS/A858, https://links.lww.com/MSS/A859, https://links.lww.com/MSS/A860, https://links.lww.com/MSS/A861, https://links.lww.com/MSS/A862). In all of the studies, the participants (n = 192 in total) were young (mean ± SD, 22.3 ± 2.7 yr), lean (body mass index [BMI] = 23.4 ± 2.2 kg·m−2), recreationally active (V˙O2 peak, n = 178, 57.8 ± 8.2 mL·kg−1·min−1) males who were metabolically healthy. All of the participants were weight stable (<2.5-kg change in body weight) for at least 3 months before experimental trials.
Exercise protocol characteristics
The exercise stimuli imposed within the studies included in this pooled analysis were homogenous, which were characterized as a single bout of moderate- to high-intensity aerobic exercise in all instances. In all trials, exercise was conducted within a controlled laboratory setting with participants exercising under the direct supervision of study experimenters. In all except one study (which involved an acute bout of swimming), the mode of exercise completed was treadmill running or ergometer cycling with indirect calorimetry (Douglas bags) used to monitor exercise intensity and determine energy expenditure and substrate oxidation (15). Across exercise trials, the intensity of exercise ranged from 56% to 83% of V˙O2 peak with a mean intensity of 69% ± 5%. The duration of each acute exercise bout ranged from 30 to 90 min (30 min, two studies; 60 min, 11 studies; 90 min, four studies).
Anthropometry and standardization
Body mass and stature were determined using standard techniques with participants wearing light clothing. Body composition (fat mass and fat-free mass) was determined using skinfold measurements (triceps, bicep, subscapular, and suprailiac) and the published equations of Durnin and Womersley (12) and Siri (35). Participants' age, stature, and body mass was used to estimate resting metabolic rate as described by Mifflin et al. (31). Participants refrained from consuming alcohol, caffeine, and participating in structured exercise for 24–48 h before main experimental trials, and during this period, dietary intake was standardized using weighed food records. Participants' last meal was consumed before study days on the prior evening (no later than 22:00), and all main trials commenced the following morning after an overnight fast. Participants maintained their habitual diet between trials in all experiments.
Hunger analyses
The primary analyses of interest in this study relating to hunger were as follows: 1) individual variation in fasting hunger (n = 192) and 2) the immediate (during exercise, n = 178) and prolonged (up to 8 h postexercise, n = 118) effects of exercise on perceived hunger. In each of the studies included within these analyses, participants reported their perceived hunger at intervals of 30 min using pen and paper–based 100-mm visual analog scales (14). The effect of exercise on hunger was assessed by comparing mean hunger ratings calculated during and after exercise with paired values calculated on each participant's control trial. In the postexercise hunger analysis, mean hunger scores were calculated from data available until the end of trials or until the occurrence of a buffet meal (when standardized appetite scores were no longer comparable). The reproducibility of fasting perceived hunger was determined from baseline hunger ratings at the start of paired exercise and control trials. Individual variation in hunger responses during and after exercise were calculated by subtracting mean hunger ratings calculated during control trials from mean hunger ratings observed during the same periods within exercise trials. For all postexercise analyses, hunger ratings obtained within the first 30 min after exercise was excluded to eliminate any latent effect of the exercise bout.
To examine the individual variation in hunger responses during and after exercise, we compared each participant's response with our new data (n = 15 young, healthy males) regarding the variation in hunger ratings across 1 h (most common duration of exercise in the present analyses) (1 h ± 30 mm; 17.2%) and over an extended duration (2.5 h ± 20 mm; 13.8%) with no intervention.
Energy intake analyses
The primary analyses of interest relating to exercise and ad libitum energy intake were 1) the effect of acute exercise on energy intake at the first meal consumed shortly after exercise (within 60 min) (n = 60) and 2) the effect of acute exercise on energy intake across several hours postexercise (range 5–9 h) (n = 128). In each of the studies included within these analyses, ad libitum energy intake was determined from buffet-style meals whereby participants had access to a range of foods for a discrete period (30 min), which was identical on paired exercise and control trials. In all trials, participants were instructed to eat until “comfortably full and satisfied” and that additional food was available if desired. All meals were consumed in isolation so that social factors did not influence eating behavior. Variation in energy intake responses to exercise was determined by subtracting each participant's energy intake during the control trial from their intake during paired exercise trials. Within the analyses examining the delayed effects of exercise on energy intake, data were included only if participants had remained in the laboratory during the entire period of observation. In addition, data were only assessed from meals consumed on the same day as exercise, i.e., data were not included from energy intake assessments conducted on the day after exercise (which occurred in three studies identified within this article).
Because the natural day-to-day variability in energy intake is highly dependent on the participants studied and the format of ad libitum meal provision (i.e., homogenous meal vs buffet meal and types of foods available at laboratory meals), we conducted a new study to characterize the variation in ad libitum energy intake across two meals (breakfast and lunch) when using a buffet meal (24, see Appendix therein) and participant cohort (n = 18; healthy, lean males) identical with that utilized within the studies described in the present manuscript. In this setting, we found that the coefficient of repeatability and intrasubject variation at breakfast was ±1937 kJ and 18.9%. Furthermore, when energy intake at breakfast was combined with a buffet lunch, together, the corresponding repeatability values were 2138 kJ and 8.9%. These boundaries of variation were used to determine the boundaries of “true variation” in energy intake responses in the present investigation.
Acylated ghrelin analyses
The primary analyses of interest relating to acylated ghrelin were as follows: 1) the immediate (during exercise, n = 118) and prolonged (up to 8 h postexercise; n = 89) effects of acute exercise on circulating acylated ghrelin concentrations and 2) the day-to-day variation in fasting circulating acylated ghrelin concentrations (n = 138). In each of the studies included within these analyses, circulating concentrations of acylated ghrelin were determined from venous blood samples taken by venipuncture (fasting measurement in one study) or cannulas (16 studies) positioned in antecubital veins. Across all studies, plasma acylated ghrelin concentrations were determined using the same enzyme-linked immune-sorbent assay (SPI-BIO, Montigney le Brettoneux, France), which has demonstrated good intraassay (typically 6%–8%) variation in our laboratory. Importantly, identical sampling pre- and posttreatment was performed across all studies as detailed previously (6). Variation in circulating acylated ghrelin responses to exercise was determined by subtracting the plasma acylated ghrelin area under the curve (AUC) during the period of interest within the control trial (exercise period and postexercise period) from the corresponding period during the exercise trial. These data were then expressed as a percentage difference with positive values indicating an increase in circulating acylated ghrelin in response to exercise (and vice versa). Acylated ghrelin data were expressed as percentage difference rather than absolute values (as per our hunger and energy intake data) because of variation in absolute acylated ghrelin values obtained across our data (most likely related to antibody variation with ELISA kits over time). To determine the day-to-day variability in circulating acylated ghrelin concentrations over an extended period, we collected new data whereby circulating acylated ghrelin concentrations were determined from six samples over a 2.5-h period on two separate days with no intervention (n = 15 healthy, young males). With diet and physical activity standardized in the prior 24 h, across a period of 1 h (the median exercise duration in the present analysis), the coefficient of repeatability and intrasubject variation for circulating acylated ghrelin was ±46 pg·mL−1 and 17.2%, respectively. Over a longer period of 2.5 h, the corresponding values were ±38 pg·mL−1·h−1 and 14.4%.
Statistical analyses
Data were analyzed using the Statistical Package for the Social Sciences version 22.0 (IBM SPSS, Inc., Chicago, IL). AUC was calculated for plasma acylated ghrelin using the trapezoidal method. Repeated-measures ANCOVA were used to assess differences in hunger (fasting and mean values), energy intake and circulating acylated ghrelin (fasting and AUC) between paired control and exercise trials. Study was included as a covariate for all analyses, whereas additional covariates were added if they correlated significantly with dependent variables. In effect, age and fat mass were included as additional covariates in the fasting hunger analyses, whereas fat mass was included as a covariate in the postexercise hunger analyses. Variation in fasting hunger ratings and circulating acylated ghrelin concentrations were expressed as the coefficient of intrasubject variation (CVintra = SDd/(m√2)) and coefficient of repeatability (CR = 2 SD) as described by Horner et al. (21). The Person product–moment correlation coefficient was used to examine relationships between key variables with the correlations interpreted as small (0.1), medium (0.3), and large (0.5) (8). Within the correlation analyses, exact participant numbers are stated in parenthesis when this deviates from the number included within the main outcome analysis. Effect sizes were calculated to determine the magnitude of statistical effects using Cohen's d, which adopts the following values to represent small (0.2), medium (0.5), and large (0.8) effects (8). All data are presented as mean ± SD. Statistical significance was identified if P < 0.05.
RESULTS
Hunger responses
Data describing paired fasting hunger scores at the beginning of an exercise and control trial were available for 192 participants (see Table, Supplementary Digital Content 1, studies included in the fasting hunger analyses, https://links.lww.com/MSS/A855). There was no significant difference in fasting hunger scores between trials (exercise, 59 ± 23 mm; control, 56 ± 24 mm; P = 0.929, d = 0.13). The intrasubject variation in fasting hunger between paired exercise and control trials was 38% with a coefficient of repeatability of ±44 mm. Fasting hunger was strongly correlated between each participant's main trials (r = 0.557, P < 0.001). Mean fasting hunger scores were positively associated with fat-free mass (n = 165, r = 0.213, P = 0.006) and age (r = 0.143; P = 0.048) and inversely related to fat mass (n = 165, r = −0.213, P = 0.006). Mean fasting hunger was not related to weight (r = −0.032, P = 0.662), BMI (r = −0.045, P = 0.537), V˙O2 peak (n =178, r = −0.057, P = 0.450), or estimated resting metabolic rate (r = −0.039, P = 0.591).
Supplementary Digital Contents 2 and 3 identify the specific studies, along with their associated characteristics, which were pooled to obtain data regarding hunger responses during (n = 178) and after (n = 118) exercise (see Table, Supplemental Digital Content, 2, Studies included in the analysis examining hunger responses during exercise, https://links.lww.com/MSS/A856; see Table, Supplemental Digital Content 3, Studies included in the analysis examining hunger responses after exercise, https://links.lww.com/MSS/A857). Mean hunger ratings during exercise were significantly lower compared with paired hunger ratings during control trials (exercise, 41 ± 26 mm; control, 61 ± 22 mm; P = 0.010, d = 0.77). Figure 1A shows each participant's net individual hunger response during exercise (difference between exercise and control) and demonstrates the wide range of responses observed (−94 to +73 mm). Notably, 79% (n = 140) of participants demonstrated suppressed hunger during exercise, whereas 19% (n = 34) documented an increase (2% showed no difference between control and exercise trials). Importantly, however, when considering the natural variation in hunger assessment with no intervention (±30 mm over 1 h), it can be seen that 37% (n = 65) of participants' hunger was suppressed to an extent greater than the boundaries of normal variation, whereas 3% (n = 5) demonstrated an increase. The remaining 60% (n = 108) lay within this boundary. Further scrutiny of these data revealed a weak inverse relationship between percent carbohydrate oxidation during exercise and mean hunger (n = 152, r = −0.177, P =0.030). There were no relationships between mean hunger during exercise and fat oxidation (n = 152, r = 0.079, P = 0.332), exercise intensity (n = 162, r = −0.100, P = 0.204), energy expenditure (n = 162, r = −0.105, P = 0.182), or V˙O2 peak (n = 164, r = −0.088, P = 0.260).
;)
FIGURE 1: Mean hunger ratings (exercise minus control) obtained during (A, n = 178) and after exercise (B, n = 118). Values above zero indicate increased hunger during or after exercise; values less than zero indicate reduced hunger. Horizontal lines represent zones of natural variation across 1 h (A, ±30 mm) and 2.5 h (B, ±20 mm).
Hunger responses after exercise were analyzed using data collected up until the end of trials, or until the provision of an ad libitum meal (range 3–8 h postexercise). There was no significant difference in mean hunger ratings after exercise between the paired exercise (44 ± 17 mm) and control trials (44 ± 18 mm) (P = 0.142, d = 0.01). Figure 1B shows the aggregate of each participant's postexercise mean hunger responses, which varied widely (−52 to +30 mm). Fifty percent (n = 59) of participants reported lower mean postexercise hunger, whereas 47% (n = 56) demonstrated higher mean postexercise hunger (3% reported no difference between trials). Importantly, when normal variation is considered, 90% (n = 106) of participants' responses lay within the boundaries of normal variation with 4% (n = 5) demonstrating higher mean hunger after exercise and 6% (n = 7) reporting lower. Within these studies, we detected a small significant correlation between postexercise hunger and fat oxidation during exercise (n = 106, r = −0.247, P = 0.011). No relationships were found between mean postexercise hunger and carbohydrate oxidation (n = 106, r = −0.011, P = 0.911), age (n = 118, r = −0.062, P = 0.504), BMI (n = 118, r = −0.055, P = 0.552), weight (n = 118, r = 0.032, P = 0.730), fat-free mass (n = 107, r = −0.081, P = 0.404), fat mass (n = 107, r = 0.082, P = 0.402), energy expenditure (n = 116, r = 0.162, P = 0.082), or exercise intensity (n = 116, r = 0.108, P = 0.250).
Energy intake responses
Data were pooled from five of our previous research studies (n = 60) to explore the diversity of ad libitum energy intake responses at one meal provided within 60 min after a single bout of moderate- to high-intensity aerobic exercise. Supplementary Digital Content 4 describes the characteristics of the individual studies included (see Table, Supplemental Digital Content 4, Studies included in energy intake analysis at the first postexercise meal, https://links.lww.com/MSS/A858). As a group, there was no significant difference in energy intake between paired exercise and control trials (exercise, 5899 ± 1778 kJ; control, 5770 ± 1966 kJ) (P = 0.977, d = 0.10) with energy intake between trials showing a strong positive correlation (r = 0.688, P < 0.001). Figure 2A shows that on a crude individual basis, there was a range of responses observed (−5005 to +4389 kJ) with 55% (n = 33) of participants consuming more and 45% (n = 27) consuming less after exercise. Importantly though, when these data are compared against the natural variation in ad libitum energy intake at one meal with no intervention (±1937 kJ; 18.9%), it is apparent that 85% (n = 51) of participants exhibited responses within this boundary of normal variation. Seven percent of participants (n = 4) documented reduced postexercise energy intake beyond this boundary, whereas 8% (n = 5) showed an increase above this boundary.
FIGURE 2: Energy intake (exercise minus control) at one meal consumed within 60 min postexercise (A, n = 60) and at multiple meals after exercise (B, n = 128). Each individual data point represents the response for a single study participant. Values above zero indicate increased energy intake after exercise; values less than zero indicate reduced energy intake after exercise. Horizontal lines represent zones of natural variation (A, ±1937 kJ; B, ±2138 kJ).
In this cohort, there was no relationship between postexercise energy intake and prior energy expenditure (r = 0.054, P = 0.720), exercise intensity (r = 0.029, P = 0.850), carbohydrate (r = 0.113, P = 0.454), or fat oxidation (r = −0.049, P = 0.746) (n = 46). Hunger ratings immediately before the first postexercise meals were lower after exercise, likely reflecting a delayed appetite suppressive effect (exercise, 59 ± 28 mm; control, 64 ± 23 mm; P = 0.006, d = 0.36). Despite this, premeal hunger did not correlate with subsequent energy intake at the first postexercise meal in the control (r = 0.158, P = 0.229) or exercise trials (r = −0.019, P = 0.886) (n = 60).
To examine the influence of acute exercise on food intake over the course of entire laboratory trial days, including multiple ad libitum meals in some instances, data from a further six studies were pooled (n =128) (see Table, Supplementary Digital Content 5, Studies included in energy intake analysis for all meals after exercise, https://links.lww.com/MSS/A859). Three of the 11 studies provided data from two ad libitum meals; the remainder utilized one meal (which was provided >1 h postexercise). As a group, there was no significant difference in energy intake between paired exercise and control trials (exercise, 9694 ± 5468 kJ; control, 9498 ± 5435 kJ; P = 0.481, d = 0.11) with responses between trials showing a strong positive correlation (P < 0.001, r = 0.949). Figure 2B shows that on a crude individual basis, there was a range of responses observed; 59% (n = 75) of participants consumed more and 41% (n = 53) consumed less after exercise. Importantly though, when these data are compared against the natural variation in ad libitum energy intake from multiple meals with no intervention (±2138 kJ; 8.9%), it is apparent that 81% (n = 105) of participants exhibited responses within this boundary of normal variation (Fig. 2B). Nine percent (n = 11) of participants documented reduced postexercise energy intake beyond this boundary, whereas 10% (n = 12) showed an increase. Across the control (r = 0.592) and exercise trials (r = 0.623), ad libitum energy intake was associated with hunger ratings (both P < 0.001) determined after exercise (or the equivalent time period on the control trial).
Acylated ghrelin responses
Data describing paired fasting acylated ghrelin plasma concentrations were available for 141 participants (see Table, Supplementary Digital Content 6, Studies included in fasting acylated ghrelin analysis, https://links.lww.com/MSS/A860). Two outliers were identified and removed from these analyses because the difference between paired samples was 4.5- and 10.5-fold greater than the SD of differences between paired samples for the cohort (±31 pg·mL−1). One additional outlier was removed because their mean fasting plasma acylated ghrelin values were 7.7 times greater than the group mean (949 vs 123 pg·mL−1). With these outliers removed (n = 138), fasting acylated ghrelin plasma concentrations did not differ between the control (125 ± 109 pg·mL−1) and exercise (121 ± 100 pg·mL−1) trials (P = 0.638, d = 0.12). The coefficient of repeatability and intrasubject variation between samples was ±63 pg·mL−1 and 19.2%, respectively. There were no significant correlations between mean fasting acylated ghrelin and hunger (r = −0.004, P = 0.959), BMI (r = −0.093, P = 0.275), weight (r = −0.091, P = 0.288), age (r = −0.015, P = 0.860), estimated resting metabolic rate (r = −0.073, P = 0.392), fat-free mass (n = 114, r = 0.092, P = 0.331), or fat mass (n = 114, r = −0.092, P = 0.331).
Acylated ghrelin responses during exercise were examined using data derived from 12 studies (n = 118; see Table, Supplemental Digital Content 7, Studies included in the analysis examining acylated ghrelin responses during exercise, https://links.lww.com/MSS/A861). In eight studies, the duration of exercise was 60 min (80 participants); in three studies, it was 90 min (30 participants); and in one study, it was 30 min (eight participants). As a group, the circulating acylated ghrelin AUC was 24% lower during exercise (99 ± 94 pg·mL−1·h−1) compared with control (131 ± 106 pg·mL−1·h−1) (P < 0.001, d = 1.0). Figure 3A shows the wide variation in acylated ghrelin responses to exercise with 89% (n = 105) of participants exhibiting lower values on their exercise trial, whereas 11% (n = 13) demonstrated higher values after exercise. Notably, when comparing these responses to the natural variation in acylated ghrelin measurement over this period (±17.2%, obtained from our new data), it can be seen that 27% (n = 32) of participants demonstrate responses, which fall within this normal range, with 66% (n = 78) and 7% (n = 8) showing a suppression and increase beyond of this range, respectively. No significant correlations were found between acylated ghrelin concentrations during exercise and exercise intensity (r = −0.111, P = 0.251) or carbohydrate oxidation (r = 0.122, P = 0.223). Fat oxidation during exercise was positively associated with acylated ghrelin concentrations (r = 0.286, P = 0.004).
FIGURE 3: Circulating acylated ghrelin concentrations (exercise minus control) during (A, n = 118) and over several hours after (B, n = 89) exercise. Each individual data point represents the response for a single study participant. Values above zero indicate increased acylated ghrelin after exercise; values less than zero indicate reduced acylated ghrelin after exercise. Horizontal lines represent zones of natural variation (A, ±17.2%; B, ±14.4%).
The prolonged effects of exercise on circulating acylated ghrelin concentrations were assessed by comparing paired postexercise acylated ghrelin AUC values across nine studies (n = 89; see Table, Supplemental Digital Content 8, Studies included in the analysis examining acylated ghrelin responses after exercise, https://links.lww.com/MSS/A862). Plasma acylated ghrelin concentrations were measured between 3 and 8 h after exercise. As a group, the postexercise acylated ghrelin AUC was 16% lower after exercise (108 ± 101 pg·mL−1·h−1) compared with control (128 ± 120 pg·mL−1·h−1) (P = 0.024, d = 0.61). Individually, Figure 3B shows that 74% (n = 66) of participants demonstrated reduced levels of acylated ghrelin, whereas 26% (n = 23) showed an increase after exercise. Notably, again, when comparing these responses with the natural acylated ghrelin sampling variation seen across an extended period (±14.4%), 42% (n = 37) of participants' responses were within the boundaries defined by this normal variation, whereas 10% (n = 9) and 48% (n = 43) of participants' responses were above and below this range, respectively.
DISCUSSION
In this study, we have pooled our research group's expansive data archive of acute experimental research trials in an effort to provide novel insights regarding the interaction between exercise and appetite regulation. Specifically, in this article, the data from 17 of our group's previous studies have been collated to interrogate interactions between exercise, hunger, ad libitum energy intake, and acylated ghrelin. Importantly, this large database of tightly controlled experimental trials has enabled us to explore intersubject variation in response to exercise, which is a key consideration in precision medicine and has begun to receive attention in energy balance research (13,18,20,38). Our findings clarify and consolidate several previously reported outcomes yet also provide new insights which have emerged from our unique collection of data.
The hunger outcomes reported here are consistent with previous findings published within and external to our laboratory, which have shown that single bouts of moderate- to high-intensity aerobic exercise transiently suppress hunger but have little effect in the hours afterward (22,23,25,26,29,30,37). Specifically, in our pool of 178 individuals, group-level analyses showed that mean hunger perceptions are suppressed by approximately one-third during exercise, which represents a medium- to large-sized statistical effect. Interestingly, there was marked variation in hunger responses, which ranged from an extensive suppression to hunger stimulation. Importantly though, even when we accounted for the natural day-to-day variation in hunger assessment that occurs when using visual analog scales, we saw that just over one-third of the study sample reported suppressed hunger below this boundary of variation, whereas only a handful of individuals reported increased hunger above this level. The remainder of participants' responses lay within the boundaries of normal variation, and therefore it is uncertain whether or not these responses represent true effects or random variation.
It is relevant to note that in our analyses, we compared our hunger data with hunger variability estimates derived from a sample of young, healthy males within our laboratory. We purposefully chose to collect these new data so that our comparator values were derived from the same population and under the same circumstances as per the experimental studies included within this manuscript. Our variability estimates showed that mean hunger can vary by ±30 mm for 1 h, which was greater than with additional assessments over a longer period of observation (2.5 h ± 20 mm). Variability estimates for hunger ratings calculated over extended durations have been published previously by others and which have ranged ±14–24 mm (14,16,21,32). These values compare favorably with ours over an extended period and support the validity of our comparisons. This new information shows that despite a large amount of variability being apparent in short-term hunger assessments, exercise is associated with a robust suppression of hunger for a large proportion of individuals. Additional work is now needed to examine whether this effect of exercise is reproducible across exposures within individuals and to identify the key moderating factors.
Our analyses of hunger responses in the hours after exercise demonstrated that single bouts of moderate- to high-intensity aerobic exercise have no effect on hunger during the remainder of the day thereafter for the majority of individuals. Again, this outcome is consistent with previous findings and confirms that acute exercise-induced energy deficits do not create an automatic drive to increase hunger (5). Notably, our data showed an even spread of net mean hunger responses postexercise; however, the vast majority of responses (90%) lay within reported boundaries of normal variation. Consequently, our data show that there is little definitive variation in postexercise hunger responses, with only 10% of individuals demonstrating changes in postexercise hunger outside the normal variation boundaries. In future studies, it would be interesting to see whether these responses are consistent across additional trials for this subset of individuals as opposed to representing random events.
Given the large number of fasting hunger ratings (n = 192) obtained at the beginning of the paired control and exercise trials, we examined the variation between repeated assessments. We identified a rather large variation in fasting hunger (38%, ±44 mm), which is consistent with results from previous studies. Specifically, in a sample of 12 active males, Gonzalez et al. (16) reported a 21% coefficient of variation, whereas in a similar population, others have calculated higher estimates (24%–30%) (32). Furthermore, Horner et al. (21) reported a higher estimate in a sample of overweight and obese males (35%). Collectively, these data identify the expected variation in fasting hunger ratings across repeated assessments in young, healthy males, and these data have implications for sample size calculations within experimental research trials. Such high coefficients of variation also support the measurement of hunger perceptions at multiple time-points in response to an intervention rather than single fasted values.
In our fasting hunger data, we identified significant, albeit weak, correlations with fat-free mass (positive) and fat mass (inverse). These findings support recent suggestions that fat-free mass is a central driver of daily food intake (4), whereas adipose tissue may exert an inhibitory effect on appetite and food intake in lean individuals (3). Homogeneity in our participants' body composition may explain the lower strength of these associations in our cohort compared with other published data (3). Alternatively, this discrepancy may be attributable to the correlational rather than causal relationships between these variables.
In our analyses, we also examined the effect of acute exercise on ad libitum energy intake at buffet meals consumed within 60 min after exercise as well as at meals consumed over several hours postexercise. Consistent with previous data collected outside our laboratory (25,26,28,33), our pooled analysis showed that at group level, energy intake was unaffected at meals consumed within the first postexercise hour. This outcome was apparent, despite hunger ratings being significantly lower (8%) immediately before ad libitum meals after exercise. Indeed, we actually found that 85% of participants' net energy intake responses (aggregate of control and exercise values) lay within the boundaries of normal day-to-day variation, as determined by our own repeatability experiment, which was conducted with a similar population and buffet meal. This is an important finding because it demonstrates that there is actually very little true variation in ad libitum energy intake beyond the summated boundaries of biological variation and technical measurement error. Previously, researchers have attempted to categorize individual participants as “compensators” or “noncompensators” with regard to the effect of exercise on energy intake based on aggregated energy intake responses after paired acute exercise and control trials (13,20). In these previous studies, it can be seen, however, that the net effect of exercise on energy intake is actually less than the natural variation in energy intake from an ad libitum meal, which has been defined as ±1406–1477 kJ (9%–12%) with ad libitum homogenous meals (17,21) and ±1937 kJ (18.9%) with ad libitum buffet meals (latter reported in this article). Moreover, a recent study has elegantly demonstrated that energy intake responses after exercise show a marked degree of inconsistency, collectively meaning that individuals cannot reliably be classified as “compensators” or “noncompensators” based on their energy intake responses to acute exercise (38). Consequently, it is likely that in our analyses, the 15% of participants who reported exercise-induced alterations in energy intake beyond normal variation boundaries may not exhibit this same response if trials were repeated.
In our energy intake analysis, it is worth noting that the identified variability estimates for our ad libitum buffet meals were considerably higher (±1937 kJ, 18.9%) than previously reported when homogenous meals are provided (17,21). This is most likely because a small change in food selection with a buffet meal on one occasion can produce large differences in energy intake across paired eating assessments. The implication of this is that for studies simply concerned with intervention effects on ad libitum energy intake, rather than food selection, a homogenous meal will reduce the variance in energy intake measurement and increase statistical power.
Our analyses are the first to examine the variation in energy intake responses to multiple meals over several hours after exercise. Again, our findings show that exercise had no effect on energy intake across this extended period. Furthermore, the vast majority of variation in responses once more lay within the boundaries of normal variation that we have determined ourselves across two ad libitum buffet meals. Our results therefore confirm previous findings demonstrating little effect of exercise on energy intake over extended periods (28) and highlight the lack of true variability in responses.
In this manuscript, we report the test–retest variability in circulating fasting acylated ghrelin concentrations, which has been calculated from a large sample of healthy males. We saw no significant difference in fasting acylated ghrelin concentrations between paired trials. This outcome supports the findings of Chandarana et al. (7), who also observed no differences in fasting or postprandial plasma acylated ghrelin concentrations, with or without dietary standardization. Despite this, in our analyses, we identified a rather large variance in fasting plasma concentrations (~19%) even with prior (24 h) dietary and physical activity standardization. This variance is composed of the technical error associated with the assay measurement (typically 6%–8% in our laboratory) and biological variation in ghrelin secretion and clearance. For the participants in these analyses, dietary standardization relied on individuals accurately maintaining and subsequently following food diaries, and it is possible that biological error could be reduced if diet is standardized for a longer period, or if participants are provided with all of their foods during the standardization phase. Future research should examine these methodological factors as it has direct relevance for appetite and gut hormone assessment in experimental appetite regulation research.
A recent meta-analysis of 18 data sets showed that acute exercise transiently suppresses circulating concentrations of acylated ghrelin with a small (Cohen's d −0.2) effect size (34). Half of the data sets from this analysis were from our laboratory, and therefore it is unsurprising that in the present analysis we identified a statistically large exercise-induced suppression of circulating acylated ghrelin during exercise. The larger effect reported in our laboratory compared with others is likely related to the characteristics of studies, particularly the exercise intensity imposed, and also to variation in assays utilized. Importantly, our data show that circulating levels of acylated ghrelin are suppressed in response to acute exercise in the vast majority of individuals examined. Of primary significance, in two-thirds of these cases, the reduction was beyond the boundaries of normal variation, which we explicitly defined for the purpose of this report. This finding highlights the consistency in the response to exercise yet poses the question of why such robust changes were not seen in the remainder of the study sample. Furthermore, the significance of this response is not fully understood and may be unrelated to appetite given that acute changes in response to exercise have not been found to be correlated consistently. In addition to this, although there have been many speculations (19), the mechanisms responsible for the exercise related perturbation of acylated ghrelin remain unclear.
In the present analysis, we identified a statistically significant reduction in circulating acylated ghrelin over the course of several hours postexercise. This finding is interesting given that on an individual study basis, a prolonged reduction in circulating acylated ghrelin in the hours after exercise has not been identified consistently. The substantially larger study sample used in this pooled analysis was therefore necessary to identify this small statistical effect. Interestingly, our data show that this persistent effect of exercise can be seen robustly in almost half of participants who exhibited suppressed ghrelin levels after exercise that were beyond the calculated range associated with normal variation. Research is now needed to identify the mechanisms producing this effect and to understand its physiological/metabolic significance.
The analyses in this article have provided a novel insight into the interaction between exercise, hunger, ad libitum energy intake, and circulating acylated ghrelin. These analyses have been made possible by the integration of over 10 yr of experimental appetite research in our laboratory using study protocols with a high degree of similarity. Our findings do however have some limitations that should be recognized. The first important consideration is the generalizability of our data. Because all of our participants were young, healthy men, we do not know whether our findings would generalize to other populations such as women, children, those who are inactive, or those who are obese. A second limitation of our data is that our homogenous sample may have inhibited the ability to identify associations between key variables reported in this article. Third, it is feasible that the energy intake response to exercise may differ between a laboratory controlled environment and an ecologically valid social setting. However, the aim of this study was to understand the physiological effects of exercise on appetite and energy intake responses in a tightly controlled laboratory environment to control against other confounding factors. Finally, it should be recognized that the studies included in the present investigation involved acute exercise protocols that commenced either in the fasted state (n = 13) or after a breakfast snack (n = 4). Although our group has shown previously that appetite and energy intake responses to acute exercise do not differ depending on feeding status (11), there is the possibility that this factor could have interacted differently across the various studies in our pooled analyses.
In conclusion, our large pooled data set confirms that single bouts of moderate- to high-intensity aerobic exercise transiently, yet robustly, suppress hunger but have no effect on ad libitum energy intake across meals consumed on the day of exercise in healthy, young men. In addition, our data show that exercise robustly suppresses circulating concentrations of acylated ghrelin, which in this novel analyses was shown to remain suppressed for several hours after exercise. Importantly, our findings underscore the necessity to consider normal day-to-day variation in these outcomes when examining variability in responses between individuals. Most notably, our research shows that in response to acute exercise, there is very little true variation in postexercise hunger and energy intake.
This research was supported by the National Institute for Health Research (NIHR) Diet, Lifestyle and Physical Activity Biomedical Research Unit based at University Hospitals of Leicester and Loughborough University. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health.
All authors declare that there are no conflicts of interest. The results of the present study do not constitute endorsement by the American College of Sports Medicine.
REFERENCES
1. Atkinson G, Batterham AM. True and false interindividual differences in the physiological response to an intervention.
Exp Physiol. 2015;100(6):577–88.
2. Bayer R, Galea S. Public health in the precision medicine era.
N Engl J Med. 2015;373(6):499–501.
3. Blundell JE, Finlayson G, Gibbons C, Caudwell P, Hopkins M. The biology of
appetite control: do resting metabolic rate and fat-free mass drive energy intake?
Physiol Behav. 2015;152(Pt B):473–8.
4. Blundell JE, Gibbons C, Caudwell P, Finlayson G, Hopkins M.
Appetite control and
energy balance: impact of exercise.
Obes Rev. 2015;16(1 Suppl):67–76.
5. Blundell JE, Stubbs RJ, Hughes DA, Whybrow S, King NA. Cross talk between
physical activity and
appetite control: does
physical activity stimulate
appetite?
Proc Nutr Soc. 2003;62(3):651–61.
6. Broom DR, Stensel DJ, Bishop NC, Burns SF, Miyashita M. Exercise-induced suppression of acylated ghrelin in humans.
J Appl Physiol (1985). 2007;102(6):2165–71.
7. Chandarana K, Drew ME, Emmanuel J, et al. Subject standardization, acclimatization, and sample processing affect gut hormone levels and
appetite in humans.
Gastroenterology. 2009;136(7):2115–26.
8. Cohen J.
Statistical Power Analysis for the Behavioural Sciences. Hillsdale (NJ): Lawrence Erlbaum Associates; 1969. p. 23.
9. Cummings DE. Ghrelin and the short- and long-term regulation of
appetite and body weight.
Physiol Behav. 2006;89(1):71–84.
10. Deighton K, Stensel DJ. Creating an acute energy deficit without stimulating compensatory increases in
appetite: is there an optimal exercise protocol?
Proc Nutr Soc. 2014;73(2):352–8.
11. Deighton K, Zahra JC, Stensel DJ.
Appetite, energy intake and resting metabolic responses to 60 min treadmill running performed in a fasted versus a postprandial state.
Appetite. 2012;58(3):946–54.
12. Durnin JV, Womersley J. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16 to 72 years.
Br J Nutr. 1974;32(1):77–97.
13. Finlayson G, Bryant E, Blundell JE, King NA. Acute compensatory eating following exercise is associated with implicit hedonic wanting for food.
Physiol Behav. 2009;97(1):62–7.
14. Flint A, Raben A, Blundell JE, Astrup A. Reproducibility, power and validity of visual analogue scales in assessment of
appetite sensations in single test meal studies.
Int J Obes Relat Metab Disord. 2000;24(1):38–48.
15. Frayn KN. Calculation of substrate oxidation rates in vivo from gaseous exchange.
J Appl Physiol Respir Environ Exerc Physiol. 1983;55(2):628–34.
16. Gonzalez JT, Veasey RC, Rumbold PL, Stevenson EJ. Consistency of metabolic responses and
appetite sensations under postabsorptive and postprandial conditions.
Appetite. 2012;59(2):228–33.
17. Gregersen NT, Flint A, Bitz C, Blundell JE, Raben A, Astrup A. Reproducibility and power of ad libitum energy intake assessed by repeated single meals.
Am J Clin Nutr. 2008;87(5):1277–81.
18. Hagobian TA, Yamashiro M, Hinkel-Lipsker J, Streder K, Evero N, Hackney T. Effects of acute exercise on
appetite hormones and ad libitum energy intake in men and women.
Appl Physiol Nutr Metab. 2013;38(1):66–72.
19. Hazell TJ, Islam H, Townsend LK, Schmale MS, Copeland JL. Effects of exercise intensity on plasma concentrations of
appetite-regulating hormones: potential mechanisms.
Appetite. 2016;98:80–8.
20. Hopkins M, Blundell JE, King NA. Individual variability in compensatory eating following acute exercise in overweight and obese women.
Br J Sports Med. 2014;48(20):1472–6.
21. Horner KM, Byrne NM, King NA. Reproducibility of subjective
appetite ratings and ad libitum test meal energy intake in overweight and obese males.
Appetite. 2014;81:116–22.
22. Howe SM, Hand TM, Larson-Meyer DE, Austin KJ, Alexander BM, Manore MM. No effect of exercise intensity on
appetite in highly-trained endurance women.
Nutrients. 2016;8:233.
23. Kawano H, Mineta M, Miyashita M, et al. Effects of different modes of exercise on
appetite and
appetite-regulating hormones.
Appetite. 2013;66:26–33.
24. King JA, Wasse LK, Broom DR, Stensel DJ. Influence of brisk walking on
appetite, energy intake and plasma acylated ghrelin.
Med Sci Sports Exerc. 2010;42(3):485–92.
25. King NA, Blundell JE. High fat foods overcome the energy expenditure induced by high intensity cycling or running.
Eur J Clin Nutr. 1995;49(2):114–23.
26. King NA, Burley VJ, Blundell JE. Exercise-induced suppression of
appetite: effects on food intake and implications for
energy balance.
Eur J Clin Nutr. 1994;48(10):715–24.
27. King NA, Hopkins M, Caudwell P, Stubbs RJ, Blundell JE. Individual variability following 12 weeks of supervised exercise: identification and characterization of compensation for exercise-induced weight loss.
Int J Obes (Lond). 2008;32(1):177–84.
28. King NA, Lluch A, Stubbs RJ, Blundell JE. High dose exercise does not increase hunger or energy intake in free-living males.
Eur J Clin Nutr. 1997;51(7):478–83.
29. Laan DJ, Leidy HJ, Lim E, Campbell WW. Effects and reproducibility of aerobic and resistance exercise on
appetite and energy intake in young, physically active adults.
Appl Physiol Nutr Metab. 2010;35(6):842–7.
30. Martins C, Morgan LM, Bloom SR, Robertson MD. Effects of exercise on gut peptides, energy intake and
appetite.
J Endocrinol. 2007;193(2):251–8.
31. Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO. A new predictive equation for resting energy expenditure in healthy individuals.
Am J Clin Nutr. 1990;51(2):241–7.
32. Raben A, Tagliabue A, Astrup A. The reproducibility of subjective
appetite scores.
Br J Nutr. 1995;73(4):517–30.
33. Schubert MM, Desbrow B, Sabapathy S, Leveritt M. Acute exercise and subsequent energy intake. A meta-analysis.
Appetite. 2013;63:92–104.
34. Schubert MM, Sabapathy S, Leveritt M, Desbrow B. Acute exercise and hormones related to
appetite regulation: a meta-analysis.
Sports Med. 2014;44(3):387–403.
35. Siri WE. The gross composition of the body.
Adv Biol Med Phys. 1956;4:239–80.
36. Stubbs RJ, Sepp A, Hughes DA, et al. The effect of graded levels of exercise on energy intake and balance in free-living men consuming their normal diet.
Eur J Clin Nutr. 2002;56(2):129–40.
37. Ueda SY, Yoshikawa T, Katsura Y, Usui T, Fujimoto S. Comparable effects of moderate intensity exercise on changes in anorectic gut hormone levels and energy intake to high intensity exercise.
J Endocrinol. 2009;203(3):357–364.
38. Unick JL, O'Leary KC, Dorfman L, Thomas JG, Strojacker K, Wing RR. Consistency in compensatory eating responses following acute exercise in inactive, overweight and obese women.
Br J Nutr. 2015;113(7):1170–7.
39. Whybrow S, Hughes DA, Ritz P, et al. The effect of an incremental increase in exercise on
appetite, eating behaviour and
energy balance in lean men and women feeding
ad libitum.
Br J Nutr. 2008;100(5):1109–15.
40. Wren AM, Seal LJ, Cohen MA, et al. Ghrelin enhances
appetite and increases food intake in humans.
J Clin Endocrinol Metab. 2001;86(12):5992–5.
