Dietary Reference Intakes for Energy (2023) / Chapter Skim
Currently Skimming:
Appendix G: Data Analysis Report
Appendix G: Data Analysis Report
Pages 277-310
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 277... ...
Roger Zoh, Associate Professor, Department of Epidemiology & Biostatistics Stephanie Dickinson, Executive Director, Biostatistics Consulting Center Lilian Golzarri Arroyo, Biostatistician II, Biostatistics Consulting Center Aaron Cohen, Biostatistician I, Biostatistics Consulting Center Jocelyn Mineo, Research Associate, Biostatistics Consulting Center 277
|
From page 278... ...
IAEA Data Preparation..........................................................................281 3. IOM Data Preparation............................................................................282 4.
|
From page 279... ...
The first task was to request data from the relevant sources, preparing Data Use Agreements (DUAs) as needed, including lists of the specific variables requested, including TEE, Basal Energy Expenditure (BEE; or basal metabolic rate [BMR]
|
From page 280... ...
Prediction equations were then developed by fitting linear models on TEE based on sex, age, PAL, weight, height, and body composition. Multiple imputation was used to estimate PAL across 20 versions of imputed data, where models were fit to each of the 20 imputations, and the results were pooled to identify final parameter estimates and standard errors (SEs) as defined by Rubin.3 Models were fit for the overall sample and were then separated by including different Body Mass Index (BMI [kg/height or length]
|
From page 281... ...
Additional documents used for the data preparation of the IAEA data: • "IAEA publications description IU 060722 - LGA notes.xlsx" -- A list of studies to include or remove as indicated by WG1 • "CLASS 2022-06-13.xlsx" -- Categories of Income by Country to indicate high-income countries for inclusion, provided by WG1 • Growth charts downloaded from https://www.cdc.gov/growth charts/index.htm: º "Weightolength_WHO.xlsx" -- Percentiles for weight-for length for infants (0–2) per World Health Organization (WHO)
|
From page 282... ...
3. IOM DATA PREPARATION IOM data were obtained by extracting data from the pdf versions of the data listed in the appendix of the IOM's 2005 publication6 and then converting into Excel.
|
From page 283... ...
TABLE 2 Inclusion and Exclusion Criteria and Sample Size for the IOM Data Set IOM Inclusion/Exclusion N Read in Table I-1 320 Read in Table I-2 525 Read in Table I-3 407 Read in Table I-4 22 Read in Table I-5 35 Read in Table I-6 319 Read in Table I-7 360 Additional Combined Pregnancy Lactation Data 382 Merge all tables together 2313 Remove participants with PAL < 1 or > 2.5 as defined in section 6.3 2283 Based on the above, two analysis-ready data files were created: 1. one including the 2,313 participants for preliminary descriptive statistics and visualizations before PAL exclusions ("IOM")
|
From page 284... ...
5. SOLNAS DATA PREPARATION DLW and physical activity data were obtained for Hispanic adults (19+)
|
From page 285... ...
data, data were kept from the main study for age and calorimeter weight, and the following variables were renamed: Weight_calorim = CSEA2 Age = CSEA3 EE_mean_kcald= CSEA4D1 EE_SD_kcald= CSEA4D2 EE_CV_kcald= CSEA4D3 The variable ‘EE_mean_kcald' was relabeled as ‘BEE.' The ethnicity for all participants in this study was coded as ‘Hispanic,' and none of the participants were pregnant or lactating. Physical activity data were also explored where 69 subjects had physical activity data from Actical.
|
From page 286... ...
The combined data set included 8,722 participants for preliminary descriptive statistics and visualizations before PAL exclusions and 8,600 observations after removing participants with PAL < 1 or > 2.5 as defined in section 6.3. Data coding and preparations were performed as follows: 6.1 Age Categories Age categories were defined as follows for descriptive statistics reports, according to "Life Stage" as indicated by WG1: • Infants are 0 to 11.99 months • Children are 12.0 months to 8.99 years • Teenagers are 9.0 to 18.99 years • Adults are 19.0 years to 101 years
|
From page 287... ...
, PAL categories were defined by WG1 accordingly, and categories (PALCAT) were calculated in the SAS code as follows: 7 Note that the term "Sedentary" and PALCAT="S" is used in this report as well as the analytic code and output, according to the labels in the 2005 IOM report before the committee relabeled the lowest level as "inactive."
|
From page 288... ...
288 DIETARY REFERENCE INTAKES FOR ENERGY If 3.0= |
From page 289... ...
1985. Predicting basal metabolic rate, new standards and review of previous work.
|
From page 290... ...
that correlated with the variable to be imputed, which improves the precision of estimates.9,10 TABLE 5 Sample Sizes for Final Analysis Data Set, after Exclusions, by Data Source Data source N IAEA 5717 IOM 2283 CNRC 220 SOLNAS 380 Combined data for analysis 8600 9 Ejima, K., R
|
From page 291... ...
1985. Predicting basal metabolic rate, new standards and review of previous work.
|
From page 292... ...
Because there were 20 imputed data sets, these values were only removed in that specific imputation, and that person would remain in the other imputations where the values remained < 2.5. 7.2 Statistical Modeling TEE models were fit separately for each strata: • Infant/Toddler Boys (0–2.99 years old)
|
From page 293... ...
Weight 0 0 1 0; estimate ‘Sedentary: Height' Height 1 PALCAT* Height 0 0 1 0; estimate ‘Low Active: Intercept' Intercept 1 PALCAT 0 1 0 0; estimate ‘Low Active: Age' Age 1; estimate ‘Low Active: Weight' Weight 1 PALCAT*
|
From page 294... ...
proc mianalyze data=Est_&agecat.; by Parameter; modeleffects Estimate ; stderr StdErr; ods output ParameterEstimates=Pooled_&agecat.; run; Children 0 to < 3 years old did not have separate models by PAL category; all data were pooled. Analysis of pregnancy data included longitudinal data for women by trimester.
|
From page 295... ...
, and ‘TEE_pred.o' is the predicted value for TEE (yˆi)
|
From page 296... ...
Parameter estimates from the TEE equations developed on the main data set were used to calculate the predicted values of TEE on the external data, and those predicted values were compared to the observed (Mean) TEE values in the external validation data using the same measures described above, such as the R-squared and Pearson correlation of observed vs predicted values, as a measure of model fit and performance.
|
From page 297... ...
APPENDIX G 297 TABLE 6 Sample Sizes for Final Analysis Data Set, by Data Source and Age Group (Appendix P §2.1) CNRC IAEA IOM SOLNAS TOTAL Infants 0 378 177 0 555 Children 0 432 689 0 1121 Teenagers 0 425 279 0 704 Adults 0 4309 767 380 5456 Preg/Lac/NPNL19 220 173 371 0 764 TOTAL 220 5717 2283 380 8600 Detailed descriptive statistics for the 8,600 observations included are presented in Appendix P, Section (§)
|
From page 298... ...
Bold numbers in Table 7 below were used to define the new age-dependent PAL categories as described above (Appendix Q §2.1)
|
From page 299... ...
TABLE 7 PAL Percentiles from Imputed Data by Age Categories (Appendix Q §2.1.1) 0 –2.99 y 3–8.99 y 9–18.99 y 19–70 .99 y 71+ y n L ac tating Pregnant Percentile n = 750 n = 926 n = 704 n = 4299 = 1281 n = 203 n = 431 10% 1.00 1.20 1.34 1.39 1.31 1.34 1.30 25% 1.11 1.31 1.50 1.53 1.46 1.50 1.46 50% 1.27 1.44 1.66 1.68 1.62 1.69 1.60 75% 1.44 1.59 1.85 1.85 1.79 1.83 1.77 90% 1.61 1.75 2.04 2.03 1.95 2.05 1.97 0–6 mo 7–11 mo 1–3 y 4–8 y 9–13 y 14–18 y 19–30 y 31–50 y 51–70 y n 71 y Percentile n = 443 n = 112 n = 243 n = 878 n = 304 n = 403 n = 1,417 n = 1,994 = 1,519 n = 1,281 10% 1.00 1.08 1.06 1.20 1.29 1.40 1.35 1.39 1.39 1.31 25% 1.07 1.19 1.17 1.32 1.44 1.56 1.50 1.53 1.52 1.46 50% 1.23 1.31 1.33 1.44 1.59 1.73 1.67 1.69 1.67 1.62 75% 1.40 1.47 1.49 1.60 1.77 1.92 1.85 1.86 1.82 1.79 90% 1.58 1.65 1.64 1.76 1.92 2.11 2.05 2.03 1.99 1.95 299
|
From page 300... ...
300 DIETARY REFERENCE INTAKES FOR ENERGY The distribution of PAL within age group is shown in Figure 3 (and Appendix Q §2.2)
|
From page 301... ...
+ 14.10 Weight (kg) • Low Active: 581.47 – 10.83 Age (y)
|
From page 302... ...
. 8.4 Model Performance Model performance and validation is outlined in the "PerformanceReport." A summary of model fit measures for the primary models including all BMI levels are listed here in Table 8 (Appendix R §1.8)
|
From page 303... ...
Weeks Predicted SE of the Strata Age Height Weight preg TEE predicted value Adult Women 19+ 53.87 162.34 71.87 -- 2,280.94 240.93 Adult Men 19+ 50.25 175.92 83.10 -- 2,930.26 342.37 Girls 3–18 9.58 135.02 37.63 -- 1,872.65 221.06 Boys 3–18 8.65 134.03 37.06 -- 2,098.77 257.61 continued 20 P
|
From page 304... ...
Tables 11A and 11B show the model fit from the predicted values after applying the TEE models to the study-level data extracted from the literature (Appendix R §4.2)
|
From page 305... ...
Boy 21 0.92 0.96 61,755.86 248.51 7.72 Girl 20 0.87 0.97 35,342.89 188.00 8.01 Man 32 0.82 0.92 49,684.74 222.90 5.57 Woman 71 0.82 0.93 28,833.66 169.80 5.46 9. APPENDICES21 Supplemental online files Appendix Description Appendix N DLW Data Codebook Appendix O Data Preparation and Preliminary Descriptive Statistics Appendix P Clean Analysis Appendix Q Multiple Imputation GLM Results Appendix R Performance Report Appendix S List of IAEA Studies with Inclusion/Exclusion continued 21 All appendixes to this IU report are provided in Supplemental Appendixes N through W and are available at: https://nap.nationalacademies.org/catalog/26818.
|
From page 306... ...
306 DIETARY REFERENCE INTAKES FOR ENERGY Appendix T IOM Data Extracted from 2002/2005 Report Appendix U External Validation Data Appendix V SAS Code for Importing, Harmonizing, and Merging Data Appendix W SAS Code for Multiple Imputation and Models
|
From page 307... ...
+ error where PALCATi represents 3 indicator variables for PAL category (Active, Low Active, Inactive) that are coded as 0 or 1; ‘A', ‘B0', ‘C0', and ‘Di' are the model coefficients for the main effects of age, height, weight and the 3 PAL categories, respectively; and ‘IBDi' and ‘IBCi' are the model coefficients for the interaction of the 3 PAL categories with height and weight, respectively.
|
From page 308... ...
The coefficients for Age, Height, and Weight may be thought of as slopes -- i.e., positive slopes represent increasing energy expenditure and negative slopes decreasing energy expenditure for a change in the corresponding variable holding the other values constant (e.g., for adult females, there is on average a decrease of 10.83 kcal/d for each 1-year increase in age, for women of the same weight, height, and physical activity level)
|
From page 309... ...
However, in contrast to the TEE prediction equation above, the intercept also remains constant, and, although the coefficients for Height and Weight vary by PAL category, they are mutltiplied by the same PA coefficient, whereas in the equation above, the parameters represent a deviation from the overall slope, which is not restricted to be the same for height and weight. A comparison of the EER values from 2005 and 2023 is presented in Chapter 7.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.