An algorithm to simulate missing data for mixed meal tolerance test response curves.

BACKGROUND: Response curves formed by analyte concentrations measured at sampled time points after consuming a mixed meal are increasingly being used to characterize responses to differing diets. Unfortunately, owing to a variety of reasons, analyte concentrations for some of the time points may be missing. OBJECTIVES: This study aimed to develop an algorithm to estimate the missing values at sampled time points in the analyte response curve to a mixed meal tolerance test (MMTT). METHODS: We developed an algorithm to simulate the missing postprandial concentration values for an MMTT. The algorithm was developed to handle any number of missing values for 2 or less consecutive missing values. The algorithm was tested on MMTT response curve data for glucose and triglyceride measurements in data from 3 different studies with 2119 postprandial MMTT response curves. The algorithm was validated by removing concentration values that were not missing and replacing them with the algorithm simulated values. The AUC error between the actual curve and simulated curves were also calculated. A web-based application was developed to automatically simulate missing values for an uploaded MMTT data set. RESULTS: The algorithm was programmed in Python and the resulting web-based application and a video tutorial were provided. The validation indicated good agreement between actual and simulated values with error increasing for less frequently sampled time points. The study with the mean minimum error of glucose concentrations was 6.2 ± 2.1 mg/dL and study with the mean maximum error of glucose concentrations was 11.3 ± 4.7 mg/dL. Triglycerides had 16.1 ± 6.2 mg/dL mean error. The AUC error was small ranging between 0.01% and 0.28%. CONCLUSIONS: The presented algorithm reconstructs postprandial response curves with estimations of values that are missing.

Transforming Big Data into AI-ready data for nutrition and obesity research.

OBJECTIVE: Big Data are increasingly used in obesity and nutrition research to gain new insights and derive personalized guidance; however, this data in raw form are often not usable. Substantial preprocessing, which requires machine learning (ML), human judgment, and specialized software, is required to transform Big Data into artificial intelligence (AI)- and ML-ready data. These preprocessing steps are the most complex part of the entire modeling pipeline. Understanding the complexity of these steps by the end user is critical for reducing misunderstanding, faulty interpretation, and erroneous downstream conclusions. METHODS: We reviewed three popular obesity/nutrition Big Data sources: microbiome, metabolomics, and accelerometry. The preprocessing pipelines, specialized software, challenges, and how decisions impact final AI- and ML-ready products were detailed. RESULTS: Opportunities for advances to improve quality control, speed of preprocessing, and intelligent end user consumption were presented. CONCLUSIONS: Big Data have the exciting potential for identifying new modifiable factors that impact obesity research. However, to ensure accurate interpretation of conclusions arising from Big Data, the choices involved in preparing AI- and ML-ready data need to be transparent to investigators and clinicians relying on the conclusions.

The impact of body composition on the degree of misreporting of food diaries.

BACKGROUND/OBJECTIVES: Accurate assessments of energy intake (EI) are needed in lifestyle interventions to guarantee a negative energy balance (EB), thereby losing weight. This study aimed (1) to compare objectively measured and self-reported EI and (2) to determine the predictors of underreporting divided by sex, adiposity and BMI category. METHODS: Seventy-three participants [mean (SD): 43.7 (9.2) years, BMI = 31.5 (4.5) kg/m2, 37% females] of the Champ4Life intervention were included in this study. EI was measured using the "intake-balance method" and self-reported through 3-day food records. Fat mass (FM) and fat-free mass (FFM) were measured by dual-energy X-ray absorptiometry. Bland-Altman analysis was performed to compare both EI assessments. RESULTS: Self-reported EI was lower than measured EI during both neutral (-355 kcal/d) and negative EB (-570 kal/day). While no significant trends were observed for EI evaluation in either neutral (p = 0.315) or negative EB (p = 0.611), limits of agreement were wide (-1720 to 1010 and -1920 to 779 kcal/day, respectively). In females, the degree of misreporting (kcal/day and %) was predicted by weight (p = 0.032 and p = 0.039, respectively) and FM (p = 0.029 and p = 0.037, respectively). In males, only BMI (p = 0.036) was a predictor of misreporting (kcal/day). CONCLUSION: Self-reported EI did not agree with measured EI. Our results show that larger body size was associated with higher levels of underestimation for EI (females only). Nevertheless, misreporting EI is a complex issue involving more associations than merely body composition. A deeper understanding could inform counseling for participants filling out food records in other to reduce misreporting and improve validity.

Exercise-induced Changes in Central Adiposity During an RCT: Effect of Exercise Dose and Associations With Compensation.

CONTEXT: Exercise can decrease central adiposity, but the effect of exercise dose and the relationship between central adiposity and exercise-induced compensation is unclear. OBJECTIVE: Test the effect of exercise dose on central adiposity change and the association between central adiposity and exercise-induced weight compensation. METHODS: In this ancillary analysis of a 6-month randomized controlled trial, 170 participants with overweight or obesity (mean ± SD body mass index: 31.5 ± 4.7 kg/m2) were randomized to a control group or exercise groups that reflected exercise recommendations for health (8 kcal/kg/week [KKW]) or weight loss and weight maintenance (20 KKW). Waist circumference was measured, and dual-energy X-ray absorptiometry assessed central adiposity. Predicted weight change was estimated and weight compensation (weight change - predicted weight change) was calculated. RESULTS: Between-group change in waist circumference (control: .0 cm [95% CI, -1.0 to 1.0], 8 KKW: -.7 cm [95% CI, -1.7 to .4], 20 KKW: -1.3 cm [95% CI, -2.4 to -.2]) and visceral adipose tissue (VAT; control: -.02 kg [95% CI, -.07 to .04], 8 KKW: -.01 kg [95% CI, -.07 to .04], 20 KKW: -.04 kg [95% CI, -.10 to .02]) was similar (P ≥ .23). Most exercisers (82.6%) compensated (weight loss less than expected). Exercisers who compensated exhibited a 2.5-cm (95% CI, .8 to 4.2) and .23-kg (95% CI, .14 to .31) increase in waist circumference and VAT, respectively, vs those who did not (P < .01). Desire to eat predicted VAT change during exercise (ß = .21; P = .03). CONCLUSION: In the presence of significant weight compensation, exercise at doses recommended for health and weight loss and weight maintenance leads to negligible changes in central adiposity.

Raising the U.S. Army Height-Weight (Body Mass Index) Standards: Quantifying Metabolic Risk.

BACKGROUND & OBJECTIVES: The U.S. Army fell 25% short of its recruitment goal in 2022 and therefore, increasing the eligibility pool for potential recruits is of interest. Raising the body mass index (BMI) standards for eligibility presents a path to increase the recruitable population; however, there may be additional costs incurred due to attendant health risks that may be present in individuals with higher BMI. METHODS: We filtered the 2017-2020 National Health and Nutrition Examination Survey by age (17-25 years) and BMI (up to 30 kg/m2). A k-means cluster analysis was performed on the filtered dataset for the variables used to determine metabolic syndrome. Metabolic syndrome Clusters were characterized through summary statistics and compared over clinical measurements and questionnaire responses. RESULTS: Five distinct clusters were identified and mean BMI in two clusters (Clusters1 and 3) exceeded the current U.S. Army BMI thresholds. Of these two clusters, Cluster 1 members had metabolic syndrome. Cluster 3 members were at higher risk for metabolic syndrome compared to members of Clusters 2, 4, and 5. Mean waist circumference was slightly lower in Cluster 3 compared to Cluster 1. None of the clusters had significant differences in depression scores, poverty index, or frequency of dental visits. CONCLUSIONS: Potential recruits from Cluster 1 have excessive health risk and may incur substantial cost to the U.S. Army if enlisted. However, potential recruits from Cluster 3 appear to add little risk and offer an opportunity to increase the pool for recruiting.

Body shape and performance on the US Army Combat Fitness Test: Insights from a 3D body image scanner.

OBJECTIVE: To identify relationships between body shape, body composition, sex and performance on the new US Army Combat Fitness Test (ACFT). METHODS: Two hundred and thirty-nine United States Military Academy cadets took the ACFT between February and April of 2021. The cadets were imaged with a Styku 3D scanner that measured circumferences at 20 locations on the body. A correlation analysis was conducted between body site measurements and ACFT event performance and evaluated using Pearson correlation coefficients and p-values. A k-means cluster analysis was performed over the circumference data and ACFT performance were evaluated between clusters using t-tests with a Holm-Bonferroni correction. RESULTS: The cluster analysis resulted in 5 groups: 1. "V" shaped males, 2. larger males, 3. inverted "V" shaped males and females, 4. "V" shaped smaller males and females, and 5. smallest males and females. ACFT performance was the highest in Clusters 1 and 2 on all events except the 2-mile run. Clusters 3 and 4 had no statistically significant differences in performance but both clusters performed better than Cluster 5. CONCLUSIONS: The association between ACFT performance and body shape is more detailed and informative than considering performance solely by sex (males and females). These associations may provide novel ways to design training programs from baseline shape measurements.

Translating digital anthropometry measurements obtained from different 3D body image scanners.

BACKGROUND: Body image scanners are used in industry and research to reliably provide a wealth of anthropometric measurements within seconds. The demonstrated utility of the scanners drives the current proliferation of more commercially available devices that rely on their own reference body sites and proprietary algorithms to output anthropometric measurements. Since each scanner relies on its own algorithms, measurements obtained from different scanners cannot directly be combined or compared. OBJECTIVES: To develop mathematical models that translate anthropometric measurements between the three popular commercially available scanners. METHODS: A unique database that contained 3D scanner measurements in the same individuals from three different scanners (Styku, Human Solutions, and Fit3D) was used to develop linear regression models that translate anthropometric measurements between each scanner. A limits of agreement analysis was performed between Fit3D and Styku against Human Solutions measurements and the coefficient of determination, bias, and 95% confidence interval were calculated. The models were then applied to normalized scanner data from four different studies to compare the results of a k-means cluster analysis between studies. A scree plot was used to determine the optimal number of clusters derived from each study. RESULTS: Correlations ranged between R2 = 0.63 (Styku and Human Solutions mid-thigh circumference) to R2 = 0.97 (Human Solutions and Fit3D neck circumference). In general, Fit3D had better agreement with Human Solutions compared to Styku. The widest disagreement was found in chest circumference (Fit3D (bias = 2.30, 95% CI = [-3.83, 8.43]) and Styku (bias = -5.60, 95% CI = [-10.98, -0.22]). The optimal number of body shape clusters in each of the four studies was consistently 5. CONCLUSIONS: The newly developed models that translate measurements between the scanners Styku and Fit3D to predict Human Solutions measurements make it possible to standardize data between scanners allowing for data pooling and comparison.

Machine learning modeling practices to support the principles of AI and ethics in nutrition research.

BACKGROUND: Nutrition research is relying more on artificial intelligence and machine learning models to understand, diagnose, predict, and explain data. While artificial intelligence and machine learning models provide powerful modeling tools, failure to use careful and well-thought-out modeling processes can lead to misleading conclusions and concerns surrounding ethics and bias. METHODS: Based on our experience as reviewers and journal editors in nutrition and obesity, we identified the most frequently omitted best practices from statistical modeling and how these same practices extend to machine learning models. We next addressed areas required for implementation of machine learning that are not included in commercial software packages. RESULTS: Here, we provide a tutorial on best artificial intelligence and machine learning modeling practices that can reduce potential ethical problems with a checklist and guiding principles to aid nutrition researchers in developing, evaluating, and implementing artificial intelligence and machine learning models in nutrition research. CONCLUSION: The quality of AI/ML modeling in nutrition research requires iterative and tailored processes to mitigate against potential ethical problems or to predict conclusions that are free of bias.

Social influences on physical activity for establishing criteria leading to exercise persistence.

Despite well-documented health benefits from exercise, a study on national trends in achieving the recommended minutes of physical activity guidelines has not improved since the guidelines were published in 2008. Peer interactions have been identified as a critical factor for increasing a population's physical activity. The objective of this study is for establishing criteria for social influences on physical activity for establishing criteria that lead to exercise persistence. A system of differential equations was developed that projects exercise trends over time. The system includes both social and non-social influences that impact changes in physical activity habits and establishes quantitative conditions that delineate population-wide persistence habits from domination of sedentary behavior. The model was generally designed with parameter values that can be estimated to data. Complete absence of social or peer influences resulted in long-term dominance of sedentary behavior and a decline of physically active populations. Social interactions between sedentary and moderately active populations were the most important social parameter that influenced low active populations to become and remain physically active. On the other hand, social interactions encouraging moderately active individuals to become sedentary drove exercise persistence to extinction. Communities should focus on increasing social interactions between sedentary and moderately active individuals to draw sedentary populations to become more active. Additionally, reducing opportunities for moderately active individuals to engage with sedentary individuals through sedentary social activities should be addressed.

Predicting energy intake in adults who are dieting and exercising.

BACKGROUND: When a lifestyle intervention combines caloric restriction and increased physical activity energy expenditure (PAEE), there are two components of energy balance, energy intake (EI) and physical activity energy expenditure (PAEE), that are routinely misreported and expensive to measure. Energy balance models have successfully predicted EI if PAEE is known. Estimating EI from an energy balance model when PAEE is not known remains an open question. OBJECTIVE: The objective was to evaluate the performance of an energy balance differential equation model to predict EI in an intervention that includes both calorie restriction and increases in PAEE. DESIGN: The Antonetti energy balance model that predicts body weight trajectories during weight loss was solved and inverted to estimate EI during weight loss. Using data from a calorie restriction study that included interventions with and without prescribed PAEE, we tested the validity of the Antonetti weight predictions against measured weight and the Antonetti EI model against measured EI using the intake-balance method at 168 days. We then evaluated the predicted EI from the model against measured EI in a study that prescribed both calorie restriction and increased PAEE. RESULTS: Compared with measured body weight at 168 days, the mean (±SD) model error was 1.30 ± 3.58 kg. Compared with measured EI at 168 days, the mean EI (±SD) model error in the intervention that prescribed calorie restriction and did not prescribe increased PAEE, was -84.9 ± 227.4 kcal/d. In the intervention that prescribed calorie restriction combined with increased PAEE, the mean (±SD) EI model error was -155.70 ± 205.70 kcal/d. CONCLUSION: The validity of the newly developed EI model was supported by experimental observations and can be used to determine EI during weight loss.

The potential epidemiologic, clinical, and economic impact of requiring schools to offer Physical Education (PE) classes in Mexico City.

BACKGROUND: Many schools have been cutting physical education (PE) classes due to budget constraints, which raises the question of whether policymakers should require schools to offer PE classes. Evidence suggests that PE classes can help address rising physical inactivity and obesity prevalence. However, it would be helpful to determine if requiring PE is cost-effective. METHODS: We developed an agent-based model of youth in Mexico City and the impact of all schools offering PE classes on changes in weight, weight-associated health conditions and the corresponding direct and indirect costs over their lifetime. RESULTS: If schools offer PE without meeting guidelines and instead followed currently observed class length and time active during class, overweight and obesity prevalence decreased by 1.3% (95% CI: 1.0%-1.6%) and was cost-effective from the third-party payer and societal perspectives ($5,058 per disability-adjusted life year [DALY] averted and $5,786/DALY averted, respectively, assuming PE cost $50.3 million). When all schools offered PE classes meeting international guidelines for PE classes, overweight and obesity prevalence decreased by 3.9% (95% CI: 3.7%-4.3%) in the cohort at the end of five years compared to no PE. Long-term, this averted 3,183 and 1,081 obesity-related health conditions and deaths, respectively and averted ≥$31.5 million in direct medical costs and ≥$39.7 million in societal costs, assuming PE classes cost ≤$50.3 million over the five-year period. PE classes could cost up to $185.5 million and $89.9 million over the course of five years and still remain cost-effective and cost saving respectively, from the societal perspective. CONCLUSION: Requiring PE in all schools could be cost-effective when PE class costs, on average, up to $10,340 per school annually. Further, the amount of time students are active during class is a driver of PE classes' value (e.g., it is cost saving when PE classes meet international guidelines) suggesting the need for specific recommendations.

Phenotypic differences between people varying in muscularity.

BACKGROUND: Body mass is the primary metabolic compartment related to a vast number of clinical indices and predictions. The extent to which skeletal muscle (SM), a major body mass component, varies between people of the same sex, weight, height, and age is largely unknown. The current study aimed to explore the magnitude of muscularity variation present in adults and to examine if variation in muscularity associates with other body composition and metabolic measures. METHODS: Muscularity was defined as the difference (residual) between a person's actual and model-predicted SM mass after controlling for their weight, height, and age. SM prediction models were developed using data from a convenience sample of 492 healthy non-Hispanic (NH) White adults (ages 18-80 years) who had total body SM and SM surrogate, appendicular lean soft tissue (ALST), measured with magnetic resonance imaging and dual-energy X-ray absorptiometry, respectively; residual SM (SMR ) and ALST were expressed in kilograms and kilograms per square meter. ALST mass was also evaluated in a population sample of 8623 NH-White adults in the 1999-2006 National Health and Nutrition Examination Survey. Associations between muscularity and variation in the residual mass of other major organs and tissues and resting energy expenditure were evaluated in the convenience sample. RESULTS: The SM, on average, constituted the largest fraction of body weight in men and women up to respective BMIs of 35 and 25 kg/m2 . SM in the convenience sample varied widely with a median of 31.2 kg and an SMR inter-quartile range/min/max of 3.35 kg/-10.1 kg/9.0 kg in men and 21.1 kg and 2.59 kg/-7.2 kg/7.5 kg in women; per cent of body weight as SM at 25th and 75th percentiles for men were 33.1% and 39.6%; corresponding values in women were 24.2% and 30.8%; results were similar for SMR indices and for ALST measures in the convenience and population samples. Greater muscularity in the convenience sample was accompanied by a smaller waist circumference (men/women: P < 0.001/=0.085) and visceral adipose tissue (P = 0.014/0.599), larger liver (P = 0.065/<0.001), kidneys (P = 0.051/<0.009), and bone mineral (P < 0.001/<0.001), and larger magnitude resting energy expenditure (P < 0.001/<0.001) than predicted for the same sex, age, weight, and height. CONCLUSIONS: Muscle mass is the largest body compartment in most adults without obesity and is widely variable in mass across people of similar body size and age; and high muscularity is accompanied by distinct body composition and metabolic characteristics. This previously unrecognized heterogeneity in muscularity in the general population has important clinical and research implications.

Food for thought: A natural language processing analysis of the 2020 Dietary Guidelines publice comments.

BACKGROUND: The Administrative Procedure Act of 1946 guarantees the public an opportunity to view and comment on the 2020 Dietary Guidelines as part of the policymaking process. In the past, public comments were submitted by postal mail or public hearings. The convenience of public comment through the Internet has generated increased comment volume, making manual analysis challenging. OBJECTIVES: To apply natural language processing (NLP NLP is natural language processing.) to identify sentiment, emotion, and themes in the 2020 Dietary Guidelines public comments. METHODS: Written comments to the Scientific Report of the 2020 Dietary Guidelines Advisory Committee that were uploaded and visible at https://beta.regulations.gov/docket/FNS-2020-0015 were extracted using a computer program and retained for analysis. All comments were filtered, and duplicates were removed. A 2-round latent Dirichlet analysis (LDA) was used to identify 3 overarching topics as well as subtopics addressed in the comments. Sentiment analysis was applied to categorize emotion and overall positive and negative sentiment within each topic. RESULTS: Three different topics were identified by LDA. The first topic involved negative sentiment surrounding removing dairy from the guidelines because the commenters felt dairy is unnecessary. The second topic focused on positive sentiment involved in restricting added sugars. The third topic was too diverse to characterize under 1 theme. A second LDA within the third topic had 3 subtopics containing positive sentiment. The first subtopic valued the inclusion of dairy in the recommendations, the second involved the health benefits of consuming beef, and the third indicated that the recommendations lead to overall good health outcomes. CONCLUSIONS: Public comments were diverse, held conflicting viewpoints, and often did not base comments on personal anecdotes or opinions without citing scientific evidence. Because the volume of public comments has grown dramatically, NLP has promise to assist in objective analysis of public comment input.

Resting Energy Expenditure: From Cellular to Whole-Body Level, a Mechanistic Historical Perspective.

The basis of heat generated by the human body has been a source of speculation and research for more than 2,000 years. Basal heat production, now usually referred to as resting energy expenditure (REE), is currently recognized as deriving from biochemical reactions at subcellular and cellular levels that are expressed in the energy expended by the body's 78 organs and tissues. These organs and tissues, and the 11 systems to which they belong, influence body size and shape. Connecting these subcellular-/cellular-level reactions to organs and tissues, and then on to body size and shape, provides a comprehensive understanding of individual differences in REE, a contemporary topic of interest in obesity research and clinical practice. This review critically examines these linkages, their association with widely used statistical and physiological REE prediction formulas, and often-unappreciated aspects of measuring basal heat production in humans.

The Behavioral Intervention with Technology for E-Weight Loss Study (BITES): Incorporating Energy Balance Models and the Bite Counter into an Online Behavioral Weight Loss Program.

This study evaluated feasibility and acceptability of adding energy balance modeling displayed on weight graphs combined with a wrist-worn bite counting sensor against a traditional online behavioral weight loss program. Adults with a BMI of 27-45 kg/m2 (83.3% women) were randomized to receive a 12-week online behavioral weight loss program with 12 weeks of continued contact (n = 9; base program), the base program plus a graph of their actual and predicted weight change based on individualized physiological parameters (n = 7), or the base program, graph, and a Bite Counter device for monitoring and limiting eating (n = 8). Participants attended weekly clinic weigh-ins plus baseline, midway (12 weeks), and study culmination (24 weeks) assessments of feasibility, acceptability, weight, and behavioral outcomes. In terms of feasibility, participants completed online lessons (M = 7.04 of 12 possible lessons, SD = 4.02) and attended weigh-ins (M = 16.81 visits, SD = 7.24). Six-month retention appears highest among nomogram participants, and weigh-in attendance and lesson completion appear highest in Bite Counter participants. Acceptability was sufficient across groups. Bite Counter use (days with ≥ 2 eating episodes) was moderate (47.8%) and comparable to other studies. Participants lost 4.6% ± 4.5 of their initial body weight at 12 weeks and 4.5% ± 5.8 at 24 weeks. All conditions increased their total physical activity minutes and use of weight control strategies (behavioral outcomes). Although all groups lost weight and the study procedures were feasible, acceptability can be improved with advances in the technology. Participants were satisfied with the online program and nomograms, and future research on engagement, adherence, and integration with other owned devices is needed. ClinicalTrials.gov Identifier: NCT02857595.

Machine learning prediction of combat basic training injury from 3D body shape images.

INTRODUCTION: Athletes and military personnel are both at risk of disabling injuries due to extreme physical activity. A method to predict which individuals might be more susceptible to injury would be valuable, especially in the military where basic recruits may be discharged from service due to injury. We postulate that certain body characteristics may be used to predict risk of injury with physical activity. METHODS: US Army basic training recruits between the ages of 17 and 21 (N = 17,680, 28% female) were scanned for uniform fitting using the 3D body imaging scanner, Human Solutions of North America at Fort Jackson, SC. From the 3D body imaging scans, a database consisting of 161 anthropometric measurements per basic training recruit was used to predict the probability of discharge from the US Army due to injury. Predictions were made using logistic regression, random forest, and artificial neural network (ANN) models. Model comparison was done using the area under the curve (AUC) of a ROC curve. RESULTS: The ANN model outperformed two other models, (ANN, AUC = 0.70, [0.68,0.72], logistic regression AUC = 0.67, [0.62,0.72], random forest AUC = 0.65, [0.61,0.70]). CONCLUSIONS: Body shape profiles generated from a three-dimensional body scanning imaging in military personnel predicted dischargeable physical injury. The ANN model can be programmed into the scanner to deliver instantaneous predictions of risk, which may provide an opportunity to intervene to prevent injury.