Metabolic characterization of meat, fish and soda intake in males: Secondary results from a randomized inpatient pilot-study

Abstract

Objective:

To evaluate whether a 12-week weight-maintaining, macronutrient stable dietary intervention that varies only by meat, fish or soda consumption alters 24hrEE and substrate oxidation.

Methods:

Healthy males were recruited to participate in a 12-week in-patient study and were randomized to a weight-maintaining dietary intervention that contained varying combinations of meat (0% or 20%), fish (0% or 6%) or soda (0% or 14%) in a factorial design. Macronutrient composition across dietary intervention groups was: 50% carbohydrate; 30% fat; and 20% protein. Twenty-four EE and substrate oxidation were measured utilizing whole-room indirect calorimetry at baseline and week 12.

Results:

Twenty-six males (data are mean ± SE) (age: 46.6 ± 10.4 years; body mass index (BMI): 26.9 ± 4.1 kg/m²); completed all measurements. Fish consumption resulted in higher 24hEE by 126 ± 55 kcals/d compared to no fish consumption (p=0.03) while 24hEE for soda consumption was 132 ± 56 kcals/day (p=0.03) lower. Approximately 80% of the decrease in 24hEE with soda consumption was due to lower awake, inactive EE (p=0.001). No specific EE component accounted for the differences observed with fish consumption.

Conclusion:

Our data indicate that dietary sources of protein and carbohydrate appear to influence 24hr EE and inactive EE.

1. Introduction

The global burden of obesity continues to persist despite public health reduction efforts and new strategies for clinical management (1). A newer body of evidence indicates that poor diet quality has deleterious health consequences including mortality, morbidity, chronic low-grade inflammation and obesity (2). Healthy dietary patterns, such as a Mediterranean-style eating pattern, are broadly described and agreed upon as those that promote physical, mental and social health and well-being and reduce overall cardiometabolic risk. Ergo, poor diet quality is the inverse and has been previously described as one containing “ultra-processed” foods derived primarily from industrially developed nutrients and additives, with little to no fruits, vegetables, whole grains or lean sources of protein (2, 3). Furthermore, consumption of ultra-processed foods and beverages, such as sugar-sweetened beverages (SSBs) and deli meats, is strongly correlated with obesogenic behaviors that support the development of cardiovascular diseases (4, 5).

Although dietary intake patterns, even within eucaloric conditions, are widely acknowledged as contributors to chronic diseases like obesity, there is very limited information available on how dietary patterns, specific foods and composition comparatively affect 24-hour energy expenditure (24hrEE), energy intake (EI) and energy balance. Factors such as macronutrient proportions and habitual caloric intake exert independent effects on ad libitum EI, body composition, 24hrEE, circulating markers of inflammation and potentially body weight (6–8). Historically, the specific effect of diet on EE has been investigated in relation to how EE is affected by total calories, macronutrient proportions or a combination of these (e.g. low fat vs. low carbohydrate diets). However, there is a lack of research evaluating the effects of macronutrient sources on 24hrEE and substrate oxidation when overall macronutrient proportions are held constant and level of processing is comparable. Therefore, the purpose of these secondary analyses from this clinical pilot study was to evaluate the effects of alterations in macronutrient sources on 24hrEE, substrate oxidation, glucose tolerance, circulating markers of inflammation and body composition while keeping macronutrient proportions stable. Using our factorial design with multiple exposures but allowing main effects analyses, we hypothesized that diets containing fish would increase 24hrEE via increases in lipid and protein oxidation, and that circulating markers of inflammation would decrease while glucose tolerance remained unchanged. For diets containing soda, we hypothesized the inverse, and for diets containing meat the investigators hypothesized that there would be no changes.

2. Methods

2.1. Participants and Eligibility Screening

This trial was approved by the National Institutes of Diabetes and Digestive and Kidney Diseases (NIDDK) Institutional Review Board (#11-DK-N018), and methods and power calculations for this randomized controlled trial have been previously described (9). Thus, the methods and data presented herein are in reference to secondary outcomes from the 12-week in-patient Developing Biomarkers of Diet (DBD) Study (NCT01237093).

Healthy males between the ages of 18–65 were recruited at the Obesity and Diabetes Clinical Research Section of the NIDDK in Phoenix, Arizona. Females were excluded from participating to eliminate risks associated with daily fish consumption in premenopausal women who may become pregnant (10). Participant eligibility was determined in two screening stages. In stage one, participants signed an informed consent, and underwent a physician-conducted health history and physical and complete metabolic panel. Participants free from acute and chronic diseases (other than obesity as defined by body mass index [BMI] of ≥ 30.0 kg/m²) subsequently entered the next round of screening. Phase two of screening was conducted on an in-patient basis, wherein all participants consumed a 3-day lead-in, weight-maintaining diet (WMD) and were instructed to remain sedentary to account for lifestyle differences. During this portion of screening, participants completed a 3-hour 75g oral glucose tolerance test (OGTT), a venipuncture, dual energy x-ray absorptiometry (DXA) scan and EE was measured over a 24-hr period utilizing a human respiratory chamber (a sample 2600 kcal chamber diet may be referenced in S1). Participants free from diabetes remained on the in-patient unit under sedentary conditions and continued the same WMD for an additional 7 days. Following the 10 collective days on the lead-in WMD, participants were then randomized (Figure 1).

Figure 1.

Study design, enrollment, and completion schematic

2.2. Experimental Design

Upon admission to the study, participants were randomized to one of eight dietary treatments. The eight diets represented all possible combinations of 3 intake variables (meat, fish and soda) at 2 levels (absent or present) in a 2³ factorial design (see Figure 1). Factorial designs allow efficient comparison of main treatment effects while allowing for multiple treatment permutations. When present, meat, fish and soda comprised 19%, 6% and 14% of energy, respectively. Participant diets were developed utilizing Food Processor (Version 11.0.2, ESHA Research). Macronutrient proportions were held constant across all eight experimental diets with 50%, 30% and 20% of energy derived from carbohydrates, fat and protein, respectively. All participants’ individualized weight-maintaining energy needs (WMEN) were calculated at baseline utilizing the metabolic research unit equation (WMEN men = 9.5 x weight (kg) + 1973) to ensure weight stability throughout the duration of the 12-week dietary intervention period, and diets were adjusted by ± 200 kcals/day to keep participants weight-stable throughout the inpatient stay (11). A one-day sample menu for day 1 of all dietary intervention combinations at a 2600 kcal level is provided for reference in the supporting information (Table S2a-h). Canned tuna and canned salmon were provided in diets that contained fish and hamburgers, hot dogs, chicken, sausage and deli meats were provided for diets that contained meat. Meat alternatives were incorporated into diets that did not contain meat and into diets that contained fish. Diets containing soda consisted of caffeinated and non-caffeinated colas; whereas, diets without soda consisted of juice and the % energy was held constant (e.g. 14%) to keep the caloric intake from liquid carbohydrates the same across diets.

2.3. Measurements and Procedures

2.3.1. Anthropometrics and Body Composition

Height and weight were measured on a digital scale and wall-mounted stadiometer. Body composition was measured via DXA (DPX-l, Lunar Corp, and Prodigy) at baseline and week-12. The use of different DXA machines was part of a cross calibration study that has been previously described. Participants were weighed daily in standardized clothing (e.g. hospital-provided gown and pants) and body weight was corrected by deducting gown and pant weight (12).

2.3.1. Glucose Tolerance and Insulin Sensitivity

A 3-hour OGTT was utilized to assess glucose tolerance and insulin sensitivity. Following an overnight fast, participants had an intravenous catheter placed and then consumed a 10-fluid ounce, 75-gram glucose beverage (Sun-Dex, Fisherbrand, Fisher Scientific, Hanover Park IL). Blood draws were completed at −15, 0, 30, 60, 120 and 180 minutes for plasma glucose and insulin. To consolidate the number of venipunctures participants underwent, plasma was also collected to measure cholesterol and inflammatory markers prior to performing the OGTT.

2.3.2. Energy Expenditure Assessment

Twenty-four hour EE, respiratory quotient (RQ), sleeping RQ, sleeping metabolic rate (SMR), awake-fed thermogenesis (AFT), inactive EE (EE0) and substrate oxidation (e.g. protein, carbohydrate and fat oxidation rates) were measured in a human respiratory chamber at baseline, (defined as pre-randomization), and at the end of the intervention (13). During baseline and follow-up 24hrEE measurements, participants were fed a diet to achieve energy balance in the confined environment of the respiratory chamber (males=1237.654 + (14.202*wt (kg))-(5.7195*BMI). EE was calculated using the Lusk equation (14). At baseline 24hrEE measurements, all participants were fed a standardized weight-maintaining chamber diet, and macronutrient proportions were 50% carbohydrate, 30% fat and 20% protein. Of note, baseline (pre-randomization) standardized chamber diets are standardized across studies conducted within the inpatient clinical research unit. During the follow-up 24hrEE measurements participants were fed foods from their respective diet intervention assignment. The morning of each 24hr EE measurement, breakfast was consumed and then participants entered the respiratory chamber. Subsequent food was provided through an air-tight interlock by trained metabolic kitchen personnel. During each chamber visit, participants were provided with a list of standardized instructions and a log to record their activities. All chambers are furnished with a bed, toilet, sink, writing desk, television and DVD player. Quality control of respiratory chamber measurements were assessed using propane with O₂ and CO₂ recoveries of 100 ± 2% on a monthly basis. If routine checks demonstrated values outside of the aforementioned recovery ranges, additional maintenance was performed until acceptable recoveries were produced.

2.3.4. Markers of Inflammation

Plasma concentrations of tumor necrosis factor alpha (TNF-α), c-reactive protein (CRP) and interleukin-6 (IL-6) were analyzed at baseline and following the intervention using the human ELISA kit (ThermoFisher Scientific). The assay range for IL-6 was 10.24 – 400 pg/mL with an intra- and inter-assay coefficient of variation (CV) of 2.2% and 9.0% respectively. For CRP, the assay range was 18.75 – 1200 pg/mL with intra- and inter-assay CV of 2.9% and 3.1% respectively. TNF-α had an assay range was 15 – 1500 pg/mL with an intra- and inter-assay CV of 10.0% and 15.1% respectively.

2.3.5. Statistical Analyses

Primary outcomes for this study were components of 24hrEE, glucose tolerance and insulin sensitivity; whereas, secondary outcomes were markers of inflammation and body composition for meat, fish and soda. Using a multivariate model with repeated measures, we had 80% power at an alpha of 0.05 to detect a 100 kcal difference in 24hrEE prior and following fish vs. no fish and soda vs. no soda, and >90% power to detect a 1mg/L change in CRP. All results are expressed as mean ± standard error. All statistical analyses were performed with SAS Enterprise Guide version 7.1. The responses of week-12 24hrEE, substrate oxidation and other metabolic variables to meat, fish and soda intake were analyzed with an analysis of covariance (ANCOVA) using a factorial design (e.g. meat versus no meat, fish versus no fish, and soda versus no soda) to increase statistical power (15). Baseline and week-12 variables for each measurement were input into an ANCOVA as the independent variable and dependent variable, respectively while meat, fish and soda were included as independent classification variables. All figures were made utilizing GraphPad Prism version 8.1.0. Additionally, to provide an overview of the raw data and the direction and magnitude of change, supporting tables (TableS3-S6) have been provided to present the delta values, or change score, and are calculated as: Week 12 – Baseline = #.

3. Results

3.1. Baseline Characteristics

Of participants enrolled, a total of 26 participants had complete measurements for baseline and follow-up inpatient respiratory chamber measurements, oral glucose tolerance tests, inflammatory marker measurements and body composition analyses. Overall, this study had an 86% retention rate and 81% of the retained sample had complete measurements. An itemized breakdown of screening, admission to the study, and retention rate are detailed elsewhere (9). Table 1 summarizes baseline characteristics of the sample in total and by treatment. Subjects’ characteristics were similar across dietary intervention groups and they were primarily Caucasians and overweight by BMI.

Table 1.

Baseline participant characteristics

Descriptive	Total	Meat	Fish	Soda
Race/ethnicity  Caucasian  American Indian  African American  Hispanic	17 7 0 2	9 4 0 0	7 4 0 2	9 5 0 1
Age (years)	46.6 ± 10.4	47.8 ± 2.0	46.8 ± 1.7	44.3 ± 2.0
BMI (kg/m2)	26.9 ± 4.1	26.8 ± 0.8	26.7 ± 0.6	25.8 ± 1.0
Weight (kg)	82 ± 12.7	84 ± 2.4	84 ± 2.3	79.6 ± 2.1
Body fat (%)	27 ± 8	27 ± 2	28 ± 1	28 ± 1
Blood pressure (mmHg)  Systolic  Diastolic	126 ± 12 79 ± 8	123 ± 2 75 ± 2	121 ± 2 74 ± 2	121 ± 2 74 ± 2
Glucose (mg/dL)	92 ± 8	93 ± 2	95 ± 2	92 ± 1
Cholesterol (mg/dL)  Total  High-density lipoprotein  Low-density lipoprotein  Triglyceride	179 ± 30 54 ± 13 107 ± 22 93 ± 59	179 ± 8 53 ± 3 110 ± 7 81 ± 8	187 ± 7 50 ± 3 115 ± 6 110 ± 17	187 ± 6 46 ± 2 118 ± 5 119 ± 15

All values are presented as mean ± standard error of the mean. Abbreviation(s) in order of appearance is/are: SEM=standard error of the mean.

3.2. Human Respiratory Chamber Measurements

At week-12, on average, 24hrEE did not change (p=0.97) in participants randomized to consume meat, but increased (126 ± 55, kcals/d, p=0.03) and decreased (−132 ± 56, p=0.03) in participants that consumed fish and soda, respectively (see Figures 2a-c and Table 2). Results were unchanged with additional adjustment for energy balance. These findings highlight that the 24hrEE differences are seen in these groups across all levels of the other dietary interventions. There were no differences in EE0 in groups that consumed meat or fish (both p>0.05); however, EE0 was lower in participants that consumed soda which accounted for approximately 80% of the change observed ((See Table 2)(p=0.001)). Figures 2a-2c present minute-by-minute energy expenditure comparing 24hrEE at baseline and week-12 for presence and absence of meat, fish and soda. No changes were observed for substrate oxidation, awake versus sleeping RQ, AFT nor any other variable measured in the respiratory chamber (all p>0.05). Mean difference (unstandardized β-coefficients) ± standard error and p-values for remaining human respiratory chamber measurements can be referenced in Table 2.

Figure 2a.

24-hour energy expenditure (24hrEE) comparisons for meat versus no meat consumption at week 12.

Figure 2c.

24-hour energy expenditure (24hrEE) comparisons for soda versus no soda consumption at week 12.

Table 2.

Mean change at week-12 for human metabolic chamber measurements

	Meat (n=13)		Fish (n=13)		Soda (n=15)
Variable	Mean ± SEM	p-value	Mean ± SEM	p-value	Mean ± SEM	p-value
24-hr EE (kcals)	−3 ± 65	p=0.97	126 ± 55	p=0.03	−132 ± 56	p=0.03
24-hr RQ (unitless)	0.02 ± 0.02	p=0.35	0.00 ± 0.02	p=0.96	0.00 ± 0.02	p=0.60
RQ Sleep (unitless)	−0.03 ± 0.02	p=0.12	.00 ± 0.02	p=0.88	−0.01 ± 0.02	p=0.79
EE Sleep (kcals)	−12 ± 59	p=0.84	103 ± 58	p=0.09	83 ± 63	p=0.20
15-hr AFT (kcals)	−22 ± 29	p=0.46	−7 ± 28	p=0.80	−34 ± 32	p=0.30
15-hr EE0 (kcals)	−43 ± 28	p=0.14	48 ± 27	p=0.90	−106 ± 28	p=0.001
CarbOx.(kcals)	−59 ± 121	p=0.63	35 ± 125	p=0.79	−113 ± 133	p=0.40
ProtOx (kcals)	−29 ± 45	p=0.52	26 ± 43	p=0.55	−10 ± 44	p=0.83
LipOx (kcals)	161 ± 128	p=0.22	47 ± 125	p=0.71	47 ± 135	p=0.73

3.3. Oral Glucose Tolerance and Insulin Sensitivity

Fasting blood glucose (FBG) did not change for meat (p=0.31), increased for fish (6 ± 2mg/dL, p=0.004) and remained unchanged for soda (p=0.94) groups. Glucose concentrations measured at 30, 60, 120 and 180 minutes during the OGTT did not change for any group (all p>0.05) in response to the intervention, and there were no changes in insulin at fasting or any other collection time points ((see Figures 3a and 3b.) (all p>0.05)). Additionally, there were no significant correlations between FBG, 24hrEE or EE0 (not shown, all p>0.05).

Figure 3.

a. Data are glucose mean ± standard deviation for meat, fish and soda at week 12..

b. Data are insulin mean ± standard deviation for meat, fish and soda at week 12..

3.4. Markers of Inflammation

Circulating plasma concentrations of IL-6, TNF-α and CRP remained unchanged for all groups following the intervention (all p>0.05). Table 3 presents mean difference (unstandardized β-coefficients) ± standard error and p-values for each group and each measurement of inflammation.

Table 3.

Mean change at week-12 for circulating markers of inflammation

	Meat (n=13)		Fish (n=13)		Soda (n=15)
Variable	Mean ± SEM	p-value	Mean ± SEM	p-value	Mean ± SEM	p-value
IL-6 (pg/mL)	0.11 ± 0.50	p=0.82	−0.20 ± 0.44	p=0.70	0.42 ± 0.44	p=0.36
TNF-α (pg/mL)	−0.62 ± 0.32	p=0.07	−0.30 ± 0.33	p=0.36	0.08 ± 0.32	p=0.81
CRP (mg/L)	−0.03 ± 0.54	p=0.95	0.57 ± 0.55	p=0.31	0.05 ± 0.54	p=0.92

All values are presented as mean difference (unstandardized β-coefficient) ± standard error of the mean. P-values < 0.05 are bolded to denote significance. Abbreviations in order of appearance is/are: SEM=standard error of the mean; IL-6=Interleukin-6; TNF-α=Tumor necrosis factor-α; CRP=C-reactive protein

3.5. Body Weight and Composition

As all participants were on a weight-maintaining diet, body weight, fat mass and fat-free mass did not vary during and following the intervention, nor did intra-individual differences in weight or body composition mediate changes in EE (all p>0.05). Table 4 presents mean difference (unstandardized β-coefficients) and p-values for each group for body composition measurements while Table S5 demonstrates the absolute differences.

Table 4.

Mean change at week-12 for body composition

	Meat (n=13)		Fish (n=13)		Soda (n=15)
Variable	Mean ± SEM	p-value	Mean ± SEM	p-value	Mean ± SEM	p-value
Body weight (kg)	−0.76 ± 0.90	p=0.41	−0.67 ± 0.91	p=0.47	0.71 ± 0.96	p=0.46
Fat-mass (kg)	−0.11 ± 0.94	p=0.91	−0.93 ± 0.95	p=0.34	0.85 ± 0.96	p=0.38
Fat-free mass (kg)	−0.54 ± 0.71	p=0.46	0.07 ± 0.71	p=0.92	0.21 ± 0.77	p=0.78
Percent body fat	0.14 ± 0.96	p=0.89	−0.82 ± 0.96	p=0.40	1.00 ± 0.96	p=0.31

All values are presented as mean difference (unstandardized β-coefficient) ± standard error of the mean. P-values < 0.05 are bolded to denote significance. Abbreviation(s) in order of appearance is/are: SEM=standard error of the mean.

4. Discussion

We conducted a long-term controlled feeding study that allowed us to perform a secondary analysis on the respective effects of macronutrient-stable WMDs containing meat, fish and soda on human energy metabolism, glucose tolerance, inflammation and body composition. We found that diets which contained fish and soda differentially influenced 24hrEE and that soda consumption was associated with a decline in EE during awake inactivity (e.g. EE0). Secondarily we observed a modest, but significant, divergence in fasting plasma glucose associated with fish. We did not find any concomitant changes in measured inflammatory markers or lipids. The differential changes in EE were observed despite no differences in dietary macronutrient proportions and occurred while participants continued a 12-week weight maintaining diet and were kept weight stable. Although participants were kept weight-stable in this study, contextually, our results demonstrate that in free-living conditions a decrease of ~135 kcals/d in 24hrEE would lead to a weight gain of 2.2 kg over 180 days when activity levels remain unchanged (16). Thus, macronutrient sources appear to influence EE.

To date, scientific consensus is that there is no unique metabolic link between SSB intake and obesity. Evidence attributes failure to adjust EI as the primary reason that SSB intake, particularly sodas, results in obesity (17–20). A review of literature published in 2010 hypothesized that the amount and source of rapidly absorbed carbohydrates, such as high-fructose corn syrup, may uniquely contribute to cardiovascular disease, type 2 diabetes and obesity (21). Although there are differences in hepatic metabolism of fructose, sucrose and high-fructose corn syrup, whether or not these may explain the changes in EE associated with soda intake observed for our study remain unknown (18). To our knowledge, there are no longer-duration randomized controlled dose response studies that have evaluated differences in EE between fruit juice and soda in humans. Results from other studies that focused on behavior modification report that fish and soda intake are associated with decreased and increased EI, respectively (22, 23). Although this study required participants to consume a weight-maintaining diet, it is worth observing that the respective 24hrEE increase and decrease for fish and soda groups inversely parallels trends reported by other studies that evaluated ad libitum EI, which may suggest that there are potentially regulatory shifts in both metabolism and appetite pathways (22, 23). Interestingly, the concomitant increase in 24hrEE for fish and decrease for soda parallel the inverse changes in body fat trends, albeit these changes were not statistically significant.

For groups whose randomization did not include soda, 100% fruit juice was substituted to keep the liquid form of calories and carbohydrates similar. For example, high-fructose corn syrup is the sole sweetener of soda in the U.S., whereas 100% fruit juices like apple juice contain a blend of glucose, fructose and sucrose that varies depending on the amount of soluble sugars the fruit source provides (24, 25). Presumably, this explains why there was not a difference when soda versus non-soda comparisons were made for RQ and whole-body substrate oxidation at week-1 and week-12, although it does not explain why EE in the inactive state significantly decreased with soda intake. Collectively, these data point to an alternate, as yet undescribed, mechanism for the lower 24hrEE findings in the soda group which may promote obesity or obesogenic behaviors. Some sodas do contain caffeine and other bioactive ingredients, but these are known to increase rather than decrease EE, as observed here. Potentially, our data provide support for the carbohydrate-insulin model of obesity, which hypothesizes that glucose-stimulated insulin secretion diverts fuels from oxidation to storage promoting increased body weight (26).

There are several potential mechanisms that pertain to why fish consumption might increase EE. The fatty acids unique to fish may affect mitochondrial respiration and skeletal muscle phospholipids, thereby altering 24hEE (27–29). Additionally, evidence indicates that intake of n-3 polyunsaturated fatty acids increases thermogenesis by promoting both the formation and activation of brown adipose tissue (30–32). Although we found significant differences in fasting plasma glucose in those exposed to fish vs. no fish, the clinical implications of this change are minimal as they are in a normoglycemic range. However, the higher fasting plasma glucose with fish consumption does implicate hepatic gluconeogenesis, which is a calorically costly metabolic process and may explain the increase in EE (33). Methionine is also lower in fish compared to meat, and methionine restriction is associated with both higher EE and lower BMI; however, this does not explain the modest but concurrent increase in FBG (34). Variance in protein digestibility (e.g. ratio of indispensable to dispensable amino acids) and turnover rate may also have impacted metabolic processes accounting for the differences between fish and other dietary protein sources included across the various diets, but previous findings on this topic are limited (35–37). Regular fish consumption resulted in higher concentrations of amino acids in some studies but other results found no differences in EE when protein quantity and type were varied (38, 39). The gut microbiota may also play a role in energy expenditure and obesity risk, and recent research in rodents and humans has suggested that the metabolites produced from polyunsaturated fatty acids promote resistance to obesity (40–42). The implications of fish intake on type 2 diabetes results remain mixed and difficult to explain, particularly in this study as there were no changes in body weight and therefore changes in fasting plasma glucose, although modest, were unexpected. Similar to ours, some studies have shown that fish intake is associated with increased risk for type 2 diabetes, while others demonstrate improvements in glucose levels (43–45).

The strengths and limitations of this study warrant further discussion. The clear strengths of our study were factorial design which allowed for a smaller sample size, the duration of the controlled feeding (12 weeks) and documented adherence to the intervention. In addition, body weight was measured daily, caloric intake was able to be adjusted to ensure body weight stability, and body composition was accounted for at both study timepoints using DXA. Whole body 24-hr EE was assessed using a human respiratory chamber and allowed for fluctuations in diurnal rhythm and pre- versus post-prandial states to be assessed, which resulted in more robust and precise measurements In terms of limitations, only 26 of our 32 participants had complete metabolic measurements for this pilot study. Additionally, although the macronutrient diet composition stayed the same, the exact foods provided in the baseline versus week-12 chambers differed. We also did not measure 24hrEE immediately following the randomization, which makes it difficult to determine whether our changes in 24hrEE are occurring acutely or only after a longer term on each diet. We also acknowledge that this was a small study population that included only men who were healthy other than obesity. This was due to concern regarding the attendant risk of rising mercury levels in females who may become pregnant, and to minimize the effect of chronic medical conditions and medications on energy metabolism. We are also unable to assess specific caffeine intake as some participants did drink caffeinated black coffee, and we are not able to specifically track the caffeine content of the ingested sodas. However, we note that if dietary caffeine was increased in the soda group, we would have expected an increase rather than a decrease in 24hrEE. We also acknowledge that although carbohydrate percentages were standardized across calorie levels, carbohydrate composition (e.g. fiber and added sugars) was not due to the differences between whole fruit juice and soda and how these items may displace other food sources. Similarly, percent energy from dietary fat was standardized, although fatty acid intake was not due to differences across interventions between meat and fish. Lastly, we do not know if different kinds of fish (e.g. lower in omega-3 fatty acids) would have affected our results.

5. Conclusions

Although obesity is now considered a pandemic, there are many aspects of its pathogenesis that remain unknown. Diet is often identified as a modifiable lifestyle factor. Therefore it is exceedingly important to continue expanding our understanding of diet to incorporate indicators of quality and account for sources of macronutrients (e.g. intake pattern) rather than relying solely upon calories and macronutrients. The results from this study suggest that macronutrient sources influence 24-hr EE and, to an extent, EE in the inactive state. The differences in protein source and quality, such as the amino acid to fatty acid ratios, along with the production source of the simple carbohydrates provided across the diets (e.g. corn for soda groups versus fruit juice from whole fruit for non-soda groups) may affect certain aspects of metabolism that are yet to be determined, but that explain our results. Our findings indicate that fish and soda differentially affect metabolic health beyond energy intake. Going forward, dietary intervention studies will need to carefully consider both sources and molecular aspects of foods provided.

Figure 2b.

24-hour energy expenditure (24hrEE) comparisons for fish versus no fish consumption at week 12.

STUDY IMPORTANCE QUESTIONS.

• Evidence indicates that dietary quality alters body weight via passive increase in energy intake (EI), but the effect on 24-hour energy expenditure (24hrEE) and substrate oxidation may also affect body weight.

• Fish and soda differentially altered 24-hour energy expenditure despite participants being kept in energy balance with standardization macronutrient proportions across diet groups.

• Unique molecular aspects of food and beverages that impact energy expenditure independent of total calories and macronutrient composition, may subsequently affect adiposity rates.