Independent effects of volume and energy density manipulation on energy intake and appetite in healthy adults: A randomized, controlled, crossover study

,


Introduction
Military personnel, endurance athletes, first responders and other populations can face challenges related to consuming enough energy to meet energy demands elicited during short-term periods of high physical activity (Burke et al., 2019;Gonzalez et al., 2022;Karl et al., 2022).Military personnel, for example, often under-consume energy during training and operations due in part to logistical constraints that limit the weight and volume of food they can carry (Karl et al., 2022) and loss of appetite (Fallowfield et al., 2014;Karl et al., 2021).Resulting energy deficits contribute to decrements in occupational performance (Murphy et al., 2018;O'Leary et al., 2020).Therefore, developing strategies to reduce energy deficits by increasing the amount of energy consumed without increasing logistical burden of weight and volume carried are of interest (Gonzalez et al., 2022;Karl et al., 2022).To optimize effectiveness, those strategies must be palatable and not suppress appetite.
The weight or volume of food that individuals consume when eating ad libitum tends to be more consistent than the amount of energy consumed (Karl & Roberts, 2014;Rolls, 2009Rolls, , 2017)).Covertly manipulating the energy density (defined as energy [kcal] per weight [g] or volume [cc]) of foods while maintaining food weight or volume is consequently an effective approach for altering short-term energy intake without impacting perceived hunger or satiety (Bell et al., 1998;Rouhani et al., 2017;Stubbs et al., 1998;Welch, 2011).The effect appears to be linear across a range of dietary energy densities (Karl & Roberts, 2014;Klos et al., 2023;Robinson et al., 2022).However, most studies have investigated foods or meals at the lower end (i.e., <3 kcal/g (Rolls, 2017)) of the possible spectrum of energy densities (0-9 kcal/g (Klos et al., 2023;Robinson et al., 2022)).Additionally, many studies have aimed to reduce intake, and therefore frame results in the context of lowering energy density as a strategy for reducing energy intake (Robinson et al., 2022) and body weight loss (Klos et al., 2023;Rolls, 2009Rolls, , 2017)).If interpreted in the opposite direction, energy density manipulation could be considered as a strategy for increasing energy intake in response to high energy demands in military environments where the weight or volume of food available is limited.Yet, few studies have examined whether increasing the energy densities of moderate-to-high energy density foods (i.e., ≥3 kcal/g) results in increased energy intake.Those that did found minor changes in total energy intake as a result (Caputo & Mattes, 1992;Cheskin et al., 2008;Miller et al., 1998).
Strategies to manipulate energy density include altering the fat, water or gas (e.g., O 2 , CO 2 ) content of foods (Klos et al., 2023;Welch, 2011;Williams et al., 2013).Most of the existing evidence base has used fat or water to adjust energy density (kcal/g).While results of those studies are generally consistent, there is some evidence that the effect size may vary according to the approach used to manipulate energy density.For example, larger reductions in total daily energy intake have been observed when energy density is lowered by decreasing fat content versus increasing water content (Williams et al., 2013).Far fewer studies have used gas to alter the volume density (g/cc) and energy density (kcal/cc) of foods (Melnikov et al., 2014;Peters et al., 2015;Welch, 2011).Those that have consistently report that reducing volume and energy densities by incorporating air into liquids (Melnikov et al., 2014;Peters et al., 2015) and foods (Osterholt et al., 2007;Rolls et al., 2000) reduces energy intake and/or appetite ratings.However, the beverages and foods used in these studies are often matched on energy content and not volume.That may confound results given that visual cues related to volume are known to impact eating behavior (Appleton et al., 2021).On the other hand, increasing volume density by removing gases through compression or rolling, and thereby increasing energy density (kcal/cc), can alter the physical structure, appearance and sensory characteristics of foods.That, in turn, may influence behavioral, physiologic and cognitive responses affecting appetite and energy intake (Appleton et al., 2021;McCrickerd & Forde, 2016;Rolls, 2017;Smeets et al., 2010).
The amount of food contained within operational ration units are limited by volume based on packaging parameters (Operational Rations, 2024) that must balance the importance of maintaining sustenance with the need to carry personal protective equipment and tactical gear.Manipulating the energy and volume densities of a food product while maintaining constant volume may therefore provide more practical solutions for minimizing energy deficits during military training and operations than increasing the volume of individual ration components.Indeed, energy dense (≥3 kcal/g), high fat (≥40% total energy), high volume density diets are being considered as one approach to overcoming this issue (Karl et al., 2022).Therefore, the primary objective of this study was to determine the independent and combined effects of energy density and volume density manipulation of foods on appetite and energy intake.

Participants
This randomized, double-blind crossover study was conducted from July 2022 to May 2023 at the U.S. Army Research Institute of Environmental Medicine.Men and women between the ages of 18-39 with a BMI ≤30.0 kg/m 2 who reported eating 3 meals/day at least 5 days/week were eligible to participate.Volunteers were instructed to abstain from a highly restrictive diet (i.e., vegan, ketogenic, Paleo), maintain body weight, avoid nicotine use, and abstain from dietary supplements known to impact appetite or metabolism throughout the study.Exclusion criteria consisted of: pregnancy/lactation; recent changes in body weight (>5 lbs); chronic medical conditions altering metabolism, appetite, or digestive physiology; inability to eat study-provided foods and beverages; use of prescription medications (other than a contraceptive) known to affect appetite, digestion and/or metabolism; and a score ≥13 for restraint or ≥9 for disinhibition on the Eating Inventory (EI) (Stunkard & Messick, 1985) scales.
The study was reviewed and approved by the U.S. Army Headquarters Medical Research and Development Command (Ft.Detrick, Fredericksburg, MD) Institutional Review Board.Investigators adhered to the policies regarding the protection of human subjects as prescribed in Army Regulation 70-25.All volunteers provided written informed consent after receiving a verbal briefing.The study was registered on clinic altrials.govas NCT05408390.

Study design
The study consisted of a baseline period followed by four 2-day treatment arms, each separated by a ≥5-day washout period.During baseline, height was measured in duplicate to the nearest 0.1 cm using a stadiometer (Seritex, Inc., Carlstadt NJ, USA) and fasted (overnight, ≥8hr) body weight was measured in light clothing using a digital scale (Taylor Precision Products, Oak Brook IL, USA) to the nearest 0.1 kg.A 3-day food and activity record was also completed.Volunteers were instructed on how to complete the food and activity records by Registered Dietitians who reviewed the records and entered the records into Food Processor SQL (v.11.11.32;ESHA Research,Salem,OR,USA).
Following baseline testing, volunteers completed four treatment arms (Fig. 1).During the first day of each treatment arm, volunteers consumed three experimental food (EXP) items, one with breakfast, one with lunch and the other as an afternoon snack.The EXP items were engineered to be low or high in energy density and low or high in volume density, creating two experimental factors: energy density (ED) and volume density (VD), and four treatment arms: low energy density + low volume density (LED/LVD), low energy density + high volume density (LED/HVD), high energy density + low volume density (HED/LVD), and high energy density + high volume density (HED/HVD).All EXP items provided during a treatment arm were of the same category (i.e., all foods were LED/LVD during one phase assigned using a random number generator to one of four treatment orders.Volunteers were instructed to abstain from strenuous activity and alcohol during each 2-day treatment arm and the day immediately before each treatment arm, and to only eat provided foods and beverages on the first day of each treatment. On the morning of the first day of each treatment, volunteers reported to the lab following an 8-hr overnight fast.Body weight was then measured, and breakfast was consumed.Volunteers then left the lab, returning for both lunch and again, for an afternoon snack.All three eating events were completed in the lab under staff supervision.Timing of eating events was kept constant within each volunteer across all four treatment arms.For each meal, volunteers were required to eat the EXP item first and completely.Immediately after the EXP item was consumed, additional meal items were then offered and consumed ad libitum (Supplemental Table 1).Appetite was measured immediately before and after, and 15 min after finishing each meal and snack.Acceptability of the EXP items was measured immediately after consuming each product.
Any uneaten meal items were provided to volunteers to consume between meals if they chose.Dinner and additional snacks were also provided for ad libitum consumption at home (Supplemental Table 1).No other foods and beverages were permitted except ad libitum water.As such, consumption of the EXP items was compulsory and any other food and beverage items provided could be consumed at the volunteer's discretion (self-selected, SSF).Total intake (TOTAL) for the test day was calculated as EXP + SSF.
Volunteers reported to the lab on the second morning of each treatment arm to return any uneaten items and trash/wrappers that were taken out of the lab on the test day, and these were reviewed and weighed by staff.Volunteers also kept a 24-hr food and activity record that day.These records were reviewed by Registered Dietitians the following day and entered into Food Processor SQL (v.11.11.32;ESHA Research, Salem, OR, USA) for analysis.A washout period of ≥5 days was then implemented before the next treatment arm (Fig. 1).To reduce bias, volunteers were informed that the study purpose was to evaluate the acceptability of new food products being developed for combat rations and were not informed that their intake was being closely monitored.

Experimental foods
For each treatment arm, three different EXP products were served, for which energy and volume densities were formulated specifically for that treatment: an egg-based breakfast bar provided at breakfast, a chicken salad sandwich at lunch, and a cheesecake for the afternoon snack.Products were formulated and manufactured by the Combat Feeding Division, DEVCOM Soldier Center (Natick, MA).Within each type of food product, served portions were equal in volume (cc) across treatment arms and the items were similar in taste, physicochemical characteristics, and appearance.Equal volume of the test products was ensured by cutting samples into the same measured dimensions.This approach maintained volume but created differences in the energy content and weight of EXP items served.Total daily energy provided from the EXP products as served ranged from ~400 kcal/d (LED/LVD) to ~1100 kcal/d (HED/HVD) (Table 1).
The volume and energy densities of the EXP items were manipulated as described below.
Egg-Bacon-Cheese Bars.Volume density was manipulated by incorporating sodium bicarbonate at 5% by weight into the LVD formulations (Supplemental Table 2).Energy density was increased by substituting cream for skim milk and whole eggs for egg whites and by slightly increasing the proportion of bacon.Bars were baked at 350 • F for min.After cooling, the loaves were trimmed with a sharp knife into cm × 5 cm x 2 cm (100 cc) blocks and packaged in trilaminate pouches.
Chicken Salad Sandwiches.Volume density was increased by compressing three slices of de-crusted white bread using a rolling pin until the height of the assembly was equal to the height of one piece of uncompressed bread (1/2 in; Supplemental Table 3).Sandwiches were thus constructed from either 2 slices of uncompressed bread or 6 slices of compressed bread.Energy density was increased by substituting full fat for low fat mayonnaise in the chicken salad mixture.Sandwiches were prepared at a bread-to-chicken salad weight ratio of 1:1 for compressed bread and 0.28:1 for non-compressed bread, trimmed to 8 cm × 8 cm lateral areas using a bakery cutter, and blast frozen at − 20F.Sandwich thickness was ~3.5 cm, providing ~225 cc volume specimens.
Lemon Cheesecake.Volume density was manipulated by incorporating sodium bicarbonate at 3.1% by weight to the LVD formulations (Supplemental Table 4).Energy density was increased by substituting high fat for fat-free cream cheese, cream for skim milk, and whole eggs for egg whites.Batter was baked at 350 • F for 25 min.Each tray was divided into three strips and vacuum microwave dehydrated for 10 min at 1 KW, after which the moisture content was ~20%.7.5 cm × 5 cm x cm (75 cc) samples were cut from the interior of the blocks, neglecting surfaces, and packaged in trilaminate pouches.
Prior to starting the study, all products were evaluated by a trained 10 member technical panel, using a 9-point hedonic scale (1-extremely poor quality; 2-poor quality; 3-moderately poor quality; 4-between moderately poor and fair quality; 5-fair quality; 6-between fair and good quality; 7-good quality; 8-between good and excellent quality; and 9-excellent quality) to judge appearance, odor, flavor, and texture attributes and overall quality.Samples were presented to the panel participants in a designated, computerized sensory laboratory, with scores collected and averaged using Sensory Information Management System (SIMS) 2000 sensory program v. 6.0.Ratings indicated that products were judged as being fair to good quality, with only minor differences across products (Supplemental Table 5).

Study diets
Individualized and standardized study diets were provided to volunteers on day 1 of each treatment arm (Supplemental Table 1).The prescribed energy content of the provided standard diet was 25% greater than estimated individual weight maintenance energy needs.Weight maintenance energy needs were determined by multiplying predicted resting metabolic rate (using sex-specific Mifflin St-Jeor equations (Mifflin et al., 1990)) by an activity factor of 1.3 (equivalent to the light activity level on test days) to account for activities of daily living and diet-induced thermogenesis.Providing energy in excess of estimated weight maintenance energy needs was done so that the combination of the EXP products and standard diet would provide more food and energy than a volunteer was likely to consume, and therefore the amount of food provided would not constrain a volunteer's total intake.The same individualized standard diet was provided across all treatment arms and contained 51%, 32%, and 18% energy from carbohydrate, fat, and protein, respectively.Diets were designed by Registered Dietitians using Food Processor SQL and consisted of commercially available foods and beverages.Intake was measured by weighing the provided and uneaten food and beverages and calculating the difference.

Appetite and acceptability
Hunger, fullness, prospective consumption (how much one could eat at that moment), and desire to eat were measured before, immediately after, and 15 min after breakfast, lunch, and afternoon snack on test days by visual analog scale.Participants indicated their sensations at that moment by marking anywhere along a horizontal 100 mm scale.The anchors of the scale were labeled with extremes of a spectrum (e.g., "not at all hungry" and "extremely hungry") (Blundell et al., 2010).Acceptability of the EXP item was measured immediately after consuming it during breakfast, lunch and afternoon snacks on test days by asking participants to rate the appearance, taste, odor, texture, and overall acceptability of the food item using a 9-pt Likert scale that ranged from 1 (dislike extremely) to 9 (like extremely).

Disinhibition, eating restraint, and variety seeking
The EI (Stunkard & Messick, 1985) questionnaire was administered to volunteers as a screening measure to determine the extent of conscious control over eating from cognitive dietary restraint and the degree of susceptibility to lose control over eating.The restraint score was calculated from 21 items (scores ranging from 0 to 21); lower scores indicated less restriction of food intake.The disinhibition score was calculated from 16 items (scores ranging from 0 to 16); higher scores were indicative of disinhibited eating (higher loss of control).Volunteers who scored ≥13 for restraint or ≥9 for disinhibition were ineligible for study participation.The 8-item VARSEEK scale (Van Trijp & Steenkamp, 1992) was administered at screening to evaluate variety seeking with respect to foods, but was not an exclusionary factor.Volunteers answered questions on a 5-point Likert scale from 1 (completely disagree) to 5 (completely agree).Higher scores indicated higher variety seeking tendencies (Van Trijp & Steenkamp, 1992).

Statistical analysis
Sample size calculations were based on expected differences in total energy intake during each test day.Total energy intake was expected to increase proportionally with covert increases in energy density (Karl & Roberts, 2014), and the magnitude of that effect was expected to be greater at higher volume densities (i.e., energy density-by-volume density interaction).Therefore, expected differences in energy intake between high ED and low ED diets from the ad libitum standard diet were expected to be ~300 kcal under high VD and ~200 kcal under low VD conditions.Total daily energy intake across phases was expected to be strongly correlated (r > 0.90 (Karl et al., 2013;O'Connor et al., 2016)) and have a standard deviation of ~750 kcal/d (O'Connor et al., 2016).Using those values, sample size calculations conducted using GLIMMPSE (www.samplesizeshop.org)and a repeated measures design with two factors (VD and ED) indicated that 20 participants would be needed to detect a VD-by-ED interaction at alpha = 0.05 and power = 0.80.
Differences in energy and macronutrient intakes across phases were assessed using linear mixed models with subject treated as a random factor, ED, VD, the ED-by-VD interaction, study phase and treatment order as fixed factors.An additional model was run for TOTAL energy in which age, baseline BMI, VARSEEK score, baseline energy intake and mean acceptability ratings were included as covariates.No covariate was associated with TOTAL energy intake and including the covariates in the model did not affect model results (data not shown).Appetite ratings were analyzed using linear mixed models with subject treated as a random factor, ED, VD, time, study phase and treatment order as fixed factors and all possible interactions between time, ED and VD added to the models.Post hoc comparisons were conducted for any significant main effects or interactions and p-values adjusted using a Bonferroni correction.Normal distribution and homoscedasticity of residuals was examined for all models and log 10 transformations were applied if needed to meet model assumptions.Differences in acceptability ratings for each food item were determined using the Friedman Test.When the Friedman test was significant (p < 0.05), differences between individual treatments and main effects of low versus high within ED and VD were examined using separate Wilcoxon Signed-Ranks Tests.Statistical analyses were performed using SPSS v.28 (IBM Inc., Armonk, New York).Data are presented as mean ± SD or median [IQR] unless otherwise noted.Significance was set at P < 0.05.

Results
Thirty-one individuals were screened for study participation.Of those screened, nine were excluded (eight due to a restraint score >13 on the EI subscales; one for dietary restrictions).A total of twenty volunteers (100% male, 22 ± 5.5 y, BMI 24.7 ± 3.7 kg/m 2 ; Table 2) were enrolled and completed the study.Participants on average scored low (4 ± 2) on the disinhibition and eating restraint (8 ± 3) subscales of the EI and displayed moderate (22 ± 7) variety seeking tendencies.No volunteer reported consuming foods or beverages other than those provided by study staff on each test day.

Dietary intake
The weight of food consumed from EXP differed across conditions (ED-by-VD interaction, p < 0.01; Fig. 2B, Supplemental Table 6).While little or no difference existed between the HED and LED conditions, a higher weight was consumed from EXP in the HVD versus LVD conditions at each level of ED (LED,HVD vs. LVD: 129 g [123,134]; HED, HVD vs. LVD: 109 g [104, 115], p < 0.01).However, no betweencondition differences in SSF or TOTAL weight consumed were observed (Fig. 2B-Supplemental Table 6).
In contrast, a main effect of ED on energy intake was observed for SSF and EXP, but not TOTAL.Energy intake from EXP was 291 kcals ([95% CI: 276,306], p < 0.01) higher during HED versus LED while energy intake from SSF was − 203 kcals (95%CI: [− 394, − 12], p = 0.04) lower.As a result, TOTAL energy intake did not differ between 282], p = 0.36) and the median percent compensation for the increase in energy intake from EXP was not different from 100% (46% [199]; one-sample Wilcoxon, p = 0.23).The difference in energy intake from EXP was 126 kcal ([95%CI: 107,144], p < 0.001) greater in the HVD versus LVD conditions relative to the HED versus LED conditions and the median percent compensation for the increase in energy intake from EXP did not differ between the VD and ED manipulations (Wilcoxon, p = 0.97).
As expected, macronutrient intake from EXP varied across conditions with ED-by-VD interactions observed for protein, fat and carbohydrate intakes (Supplemental Table 6).No ED-by-VD interactions were observed for TOTAL intake or intake from SSF for any macronutrient.However, a main effect of VD was observed for fat intake from SSF and TOTAL fat intake wherein SSF fat intake was 8 g ([95%CI 1, 15], p = 0.03) lower during HVD versus LVD but TOTAL fat intake was 11 g ([95% CI: 4, 19], p < 0.01) higher.A main effect of ED on TOTAL fat intake was also observed with intake being higher during HED versus LED (23 g [95% CI: 16, 30], p < 0.01).(Fig. 2D-Supplemental Table 6).Finally, a main effect of ED was observed for protein intake from SSF (12 g [95% CI: 1, 23], p = 0.04) and TOTAL (15 g [95% CI: 4, 26], p = 0.01) wherein intake was lower during HED versus LED for both (Fig. 2C-Supplemental Table 6).No interactions or main effects of ED or VD on energy or macronutrient intake during the 24 h period after each test day were observed (Supplemental Table 6).

Appetite
Changes in appetite ratings over time were largely unaffected by diet with no significant 3-way or 2-way interactions observed for any metric (Fig. 3A-D).A main effect of ED was observed for hunger wherein mean hunger scores were 6% higher during HED versus LED (p = 0.02) (Fig. 3A).No main effects of ED or VD were observed for fullness, prospective consumption or desire to eat.

Acceptability
Median appearance, odor, taste, texture and overall acceptability ratings ranged from 5 ("neither like nor dislike") to 8 ("like very much") for all EXP products (Supplemental Table 7).Ratings within each category generally did not differ across conditions for the sandwich and cheesecake.The one exception was that the HED sandwiches were rated as tasting better than the LED sandwiches (p = 0.04).Appearance, texture, and taste ratings for the egg bars varied across conditions and overall acceptability of the HVD egg bars was rated higher than the LVD egg bars (Supplemental Table 7).

Discussion
Consuming enough energy to meet high energy demands can be challenging for military personnel wherein logistical constraints limit the volume of food that can be carried.This study examined the independent and combined effects of manipulating the energy-and volumedensities of moderate energy density foods on energy intake and appetite to determine whether these manipulations may encourage increased energy intake in situations where food volume may be a limiting factor.Results failed to show any interaction between energy and volume density manipulation but did reveal distinct independent effects of each approach.Specifically, increasing the volume densities of isovolumetric compulsory food items through physical compression or not leavening products failed to elicit a compensatory reduction in ad libitum energy intake, resulting in higher total daily energy intakes.In contrast, increasing the energy densities of isovolumetric compulsory food items primarily by substituting higher-fat for lower-fat ingredients elicited a compensatory reduction in ad libitum energy intake that ultimately prevented any increase in total daily energy intake.
The absence of a compensatory response in appetite following compulsory intake of the higher volume density, higher energy foods is consistent with the small number of studies that have likewise manipulated volume densities of foods using air.Studies have reported higher ad libitum energy intake from a non-aerated compared to aerated high energy density snack when isovolumetric portions were served (Osterholt et al., 2007).Further, greater post-prandial hunger (Melnikov et al., 2014;Peters et al., 2015;Rolls et al., 2000) and energy intake (Rolls et al., 2000) have been observed when consuming isocaloric, lower-volume, non-aerated versus higher volume aerated food/beverages (Melnikov et al., 2014;Osterholt et al., 2007;Peters et al., 2015;Rolls et al., 2000).Novel aspects of the current study are that the manipulated foods were compulsory and, while differing in volume density, were matched on volume and provided at two levels of moderate energy density.Thus, these findings confirm and extend the evidence base by demonstrating that reducing food volume can facilitate higher energy intakes and the effect is seen across studies of different designs and persists at moderate energy densities.This effect of food volume on appetite and energy intake may be due to the inverse association between stomach distension and appetite, potentially caused by gastric mechanoreceptor activation (Geliebter et al., 1988;Khan et al., 1993;Oesch et al., 2006), or a cognitive effect of perceived volume on expected satiety (Brunstrom et al., 2010).
Whether food volume was more tightly regulated than weight or energy consumed in this cohort was unclear.The volume density manipulation resulted in compulsory items that while equal in volume, increased in energy content and weight as volume density increased.Yet, self-selected intakes between the high-and low-volume density conditions did not differ in energy content or weight consumed.Those observations may indicate that the volume of food consumed did not differ between high-and low-VD conditions, although that was not directly measured.In contrast, energy density manipulation by increasing fat content resulted in compulsory food items differing in energy content but not weight.Participants responded by reducing selfselected food intake and demonstrating a tendency to reduce the weight of self-selected foods consumed.That compensatory response was unexpected, and while consistent with some studies (Bell & Rolls, 2001), is contradictory to most of the evidence base (Klos et al., 2023;Robinson et al., 2022;Rolls, 2009Rolls, , 2017;;Rolls et al., 1999) reporting linear Fig. 2. Nutrient intake per intervention group.All food was provided and portions determined from an assigned calorie level based on energy requirements.Bars are mean +SD.SSF, self-selected foods = nutrients consumed from ad libitum food items.EXP, experimental foods = nutrients from food that was required to be consumed.TOTAL = EXP + SSF intake.HED, high energy density; LED, low energy density; HVD, high volume density; LVD, low volume density.relations between energy density and energy intake (Karl & Roberts, 2014;Klos et al., 2023;Robinson et al., 2022).
Several factors could explain this unexpected result.First, this study manipulated moderate-to-high ED foods whereas previous studies have manipulated foods having a lower energy density, commonly <2.0 kcal/ g (Klos et al., 2023;Robinson et al., 2022).Possibly, the effects of energy density and/or fat manipulation on intake are attenuated in foods having higher energy densities.That possibility was identified in a recent systematic review and meta-analysis, which also called for additional research on the effects of energy density manipulation of higher-energy density foods on energy intake (Robinson et al., 2022).Second, some evidence suggests that using fat to manipulate preload energy density rather than water or fiber may lead to greater, albeit still incomplete, energy intake compensation at the subsequent eating event (Rouhani et al., 2017).Thus, the use of fat to manipulate energy density in this study may have attenuated effects of energy density manipulation on energy intake.This also aligns with studies demonstrating that energy density is more impactful than fat content on total energy intake (Bell et al., 1998;Raben et al., 2003;Rolls et al., 1999;Saltzman et al., 1997;Stubbs et al., 1996;van Stratum et al., 1978).Third, effects of energy density manipulation on energy intake are lessened when only a portion of the diet is manipulated, as was done herein, compared to manipulating the full diet (Robinson et al., 2022).Fourth, the ability for participants to eat, and therefore compensate for energy density manipulation, throughout the day may have enabled more precise energy compensation given that longer time intervals between manipulated meals or preloads has been associated with less precise energy compensation (Almiron-Roig et al., 2013).Finally, population characteristics may have contributed as all study participants were physically active young adults.Both younger age (Appleton et al., 2011;Rolls et al., 1995) and habitual exercise (Long et al., 2002) have been associated with more precise compensation to experimental manipulations of energy intake.
Perceived appetite was generally not affected by energy and volume density manipulation, despite differences in energy intake from experimental foods and/or total energy intake, which is consistent with the evidence base.The one exception was hunger ratings, which were marginally higher during the HED relative to LED conditions.The finding that hunger was slightly higher during HED conditions while self-selected energy intake was higher during LED conditions could suggest that participants did not fully act on their hunger during HED conditions.That inaction may be another factor contributing to the absence of differences in total energy intake between the high-and lowenergy density conditions.Why this would occur is uncertain, and it is not clear whether participants would have chosen to eat more during the HED conditions had their menu not been restricted.Participants did selfselect higher protein intakes during the LED versus HED conditions.As protein is more satiating than fat or carbohydrate (Paddon-Jones et al., 2008), that choice could have contributed to differences in hunger ratings.On the other hand, the difference in hunger ratings between HED and LED conditions was only 6%, which may not have been a strong enough signal to impact eating behavior (Blundell et al., 2010).
Several limitations should be considered when interpreting study results.The HVD breakfast item formulations were rated more favorably than the LVD formulations.That difference could have contributed to the lack of differences in ad libitum energy intake between the HVD and LVD conditions.However, this seems unlikely as any effects of acceptability on appetite are unlikely to persist throughout the entire day and acute intake is impacted more under conditions of high palatability (Yeomans et al., 2004).Additionally, acceptability ratings were not associated with TOTAL energy intake (p = 0.29) and adjusting for acceptability ratings in the analysis for TOTAL energy intake did not alter results (P ED*VD = 0.85, P ED = 0.36, P VD = 0.01; data not shown).Given that the sample size in this study was relatively small and the Fig. 3. Self-reported ratings of A) hunger, B) fullness, C) prospective consumption, and D) desire to eat before and after meals/snack consumed in the laboratory on testing days.All food was provided and portions determined from an assigned calorie level based on energy requirements.Bars are mean ± SD.HED, high energy density; LED, low energy density; HVD, high volume density; LVD, low volume density.
A. Hatch-McChesney et al. cohort homogeneous, the generalizability of the results may be limited and further investigation in larger, more diverse cohorts is warranted.The short intervention duration is also a limitation, and it is not clear whether observed effects would persist over a longer period.An additional consideration regarding the experimental design is that food consumption early in the day may have impacted intake at later eating occasions.While that design may be more generalizable than single meal studies, it did restrict the focus of the analyses to cumulative intakes over the full day rather than at individual meals.Additionally, a small proportion of food items were manipulated and it is not clear if effects would be more pronounced or attenuated if the entire diet or a greater proportion of the diet was manipulated.Lastly, the provided diet was designed to offer diverse food choices but was still limited in selection compared to what participants would normally be able to access, and that could have attenuated compensatory behavior.However, that possibility seems unlikely as compensation was observed during the HED versus LED conditions.Further, the restricted menu provided a stronger level of experimental control and is more generally applicable to military settings wherein food selection from rations is limited.

Conclusion
Study findings demonstrate that increasing volume densities of a small portion of the diet through physical compression and not incorporating air can promote increased energy intake without impacting appetite and that the effects of energy density manipulation on energy intake may not be linear at moderate-to-high energy densities.Findings are particularly relevant to military personnel engaged in training and operations wherein high energy intakes are required to meet energy demands and the volume of food that can be carried is limited.Within that context, removing air/gas may be a feasible approach to promote higher energy intakes and possibly more effective than increasing fat content.Whether observed effects persist when all, or a larger portion, of the diet is manipulated requires investigation in larger and more diverse study cohorts.

Table 1
Nutrient composition of experimental food items.Values are mean ± SD; HED, high energy density; LED, low energy density; HVD, high volume density; LVD, low volume density; CHO, carbohydrate.
a Breakfast item (egg bacon bar).bLunch item (chicken salad sandwich).cAfternoon snack item.dA.Hatch-McChesney et al.

Table 2
Baseline demographic characteristics of study population.
a Data missing for 1 volunteer; Values are means ± SD or n (%).A.Hatch-McChesney et al.