3

Methodological Approach to Gathering Evidence

The committee’s approach to gathering evidence since the first Dietary Reference Intakes (DRIs) for energy published in 2005 was to conduct an umbrella review—a review of existing systematic reviews relevant to the questions in the statement of task. In the absence of a de novo systematic review relevant to the study task, the committee identified topics considered to be of highest priority and carried out the umbrella review supplemented with search terms to identify publications of randomized controlled trials and prospective longitudinal trials from the peer-reviewed published literature. This evidence-gathering approach differs from the approaches used for the two previous DRI updates (i.e., vitamin D and calcium; sodium and potassium), which included new or updated existing systematic reviews as the primary evidence to support committee deliberations, and it differs from the narrative literature review used in the first DRIs for energy, which the committee determined would not support the range or the quality of evidence needed to carry out its task.

The committee’s methodological approach included both a new analysis of doubly labeled water (DLW) data and a review of evidence on energy expenditure. For the latter, the committee considered relevance of the assessment methodology used and its applicability to the general U.S. and Canadian populations, including age, sex, body size, body composition, physiological state (e.g., pregnancy and lactation), and level of physical activity consistent with normal health. Because common physiological states such as underweight, overweight/obesity, and prediabetes/diabetes are known to influence energy requirements, the committee sought

evidence for relationships between energy balance/imbalance states and health outcomes related to the risk of chronic disease.

APPROACH FOR GATHERING DOUBLY LABELED WATER DATA

The committee was tasked to use data from studies that measure energy expenditure using DLW, which is considered the benchmark standard, to use in updating the Estimated Energy Requirement (EER) equations. Specifically, the committee was directed to use the International Atomic Energy Agency (IAEA) DLW database to derive the total energy expenditure (TEE) equations, which are the basis for deriving the EER equations. The committee augmented this database with data from other DLW studies of diverse populations known to be missing in the IAEA database. These additional data sources included the prior DRI report for energy (IOM, 2002/2005); the Study of Latinos: Nutrition and Physical Activity Assessment (SOLNAS) from the National Heart, Lung and Blood Institute (NHLBI) Biologic Specimen and Data Repository; and a source of data on pregnant and lactating women from the Children’s Nutrition Research Center at the Baylor College of Medicine (CNRC).

The Indiana University School of Public Health-Bloomington was contracted to perform statistical analysis to analyze the DLW data and generate prediction equations for TEE by age/sex and life-stage groups. This analysis was then used by the committee to derive its equations for energy expenditure by population age/sex and life-stage group. A report of this statistical analysis is in Appendix G, and the committee’s application of the analysis to derive EER equations is in Chapter 5.

APPROACH AND PROCESS FOR GATHERING RELEVANT LITERATURE

To identify the systematic reviews to include in its umbrella review, the committee developed a list of key topics and questions (Table 3-1) relevant to its statement of task and determined which topics required searches for existing systematic reviews. The committee then defined search terms (Appendix D) and eligibility criteria (Appendix E) for the studies included in the existing systematic reviews, and conducted literature searches in PubMed, Web of Science, and Embase between February and June 2022.

Generally, existing systematic reviews were eligible if they were published since 2000 (in English), evaluated studies of humans, and included studies that were comparative (comparing different interventions or exposures) or provided multivariable regressions of exposures of interest. Excluded were reviews of populations or samples that are not representative of generally healthy individuals in the United States or

Canada, including studies of people with specific health conditions (other than diabetes or weight category) or atypical metabolic states (e.g., athletes), as well as reviews of studies conducted in low-income or lower-middle-income countries. Additional systematic reviews and primary studies identified from other sources (e.g., reference lists, studies known to committee members, studies found in searches for other key questions) were also reviewed for possible inclusion. Eligibility criteria for these searches are listed in Appendix E.

For some key questions, no relevant systematic reviews were found. In these cases, additional informal literature searches of the primary peer-reviewed literature were conducted.

For each literature review, all search results were uploaded to a systematic review management program (Covidence) for screening. Two independent reviewers screened each article title and abstract. At least one of the reviewers was a committee member; the other reviewer was another committee member or a trained staff member. Any conflicting results were discussed by the reviewers and resolved by consensus. The full texts of included articles were reviewed by a single reviewer to determine eligibility for data extraction based on the predefined eligibility criteria. The questions used in the search were formulated by the committee and defined relevant to the task. Table 3-1 provides the results for each question reviewed.

All relevant data were extracted from each eligible systematic review into a customized spreadsheet. The extracted data included general information about the article, demographic characteristics (e.g., sample population age, sex, number of participants, life stage, race, and ethnicity), the number of studies, the specific intervention examined, the outcome, quantitative and narrative summaries of results, and information on the risk of bias of the included studies.

During the data extraction process, the methodological quality of each systematic review was evaluated using the Assessment of Multiple Systematic Reviews 2 (AMSTAR 2)¹ quality assessment tool, with minor adaptations for clarity (see Appendix F). The AMSTAR 2 questions were answered for each article by the committee member reviewing the key question. The answers were reviewed by a committee member with expertise in systematic reviews, who provided primary input on questions related to literature search strategy, risk-of-bias assessment, and meta-analytic method. This committee member also assigned an overall assessment of the quality of each review (as described in Appendix F). Revisions and overall assessments were then reviewed by the original data extractor and disagreements were discussed and settled by consensus.

___________________

¹https://amstar.ca/Amstar-2.php (accessed November 11, 2022).

TABLE 3-1 Evidence Map: Literature Search and Screening Results

^a During the data extraction stage, 2 articles (systematic reviews) were moved into this review from other reviews and 7 additional articles were identified by the committee from the primary literature.

^b During the title/abstract screening stage, 2 articles were moved into this review from other reviews.

^c For this search, 317 articles were identified in the database searches. After removing 19 duplicates, 298 articles remained. An additional 23 articles were identified by the committee during title/abstract screening, resulting in a total of 321 articles for screening.

^d Five additional articles were identified by the committee during data extraction.

^e Three additional articles were identified by the committee during the full text screening stage.

^f One article was moved into this review from another review during the data extraction stage.

^g Two additional articles were identified by the committee during the full-text screening stage.

Following extraction, data synthesis was conducted separately for each of the key questions. The committee did not conduct de novo meta-analyses or other reanalyses of data. Reported risk of bias analyses and GRADE (or equivalent) certainty of evidence conclusions were considered from the existing systematic reviews. To document its search results, the committee developed summary tables of the studies identified based on relevant indicators used to assess DRI values (Appendix J). These included study design, setting, participant age, biological sex, life-stage group, physiological state, intervention or exposure, methods used to determine nutrient intake levels and outcome measures, sample size, attrition, and a description of the statistical analysis. Finally, the committee identified research gaps in the process of deriving DRIs for energy and in evaluating the public health implications (Chapter 9).

REFERENCES