2
Methodological Approach to Evidence Scanning
In response to its charge, the Committee on Scanning for New Evidence on the Nutrient Content of Human Milk held meetings and discussions to develop a systematic approach to searching and summarizing relevant publications on the nutrient content of human milk and the consumption volume of human milk, as indicators of infant nutritional requirements, which will serve as a basis for developing intake recommendation through the Dietary Reference Intake (DRI) process. The committee describes below its approach to the task and the strategy used to carry out the evidence scan.
JUSTIFICATION
The U.S. Department of Agriculture (USDA) National Nutrient Database for Standard Reference (USDA SR Database) (USDA, 2019) and the Canadian Nutrient File (Health Canada, 2015) are the reference food composition databases in the United States and Canada, respectively. These databases serve as a reference standard for developing nutrient intake recommendations for infants up to 12 months of age. The nutrient profiles for human milk found in the USDA SR Database are now almost 40 years old and based on limited data, and the data in the Canadian Nutrient File were derived from the USDA SR Database. Given that new or updated DRI reviews require a systematic review, the usefulness of the USDA SR Database as a data source is limited.
Human milk nutrient profiles are still available, but they are not intended for estimating current exposures. Thus, adequate published data on the macro- and micronutrient content of human milk is needed to update the food composition databases to support systematic reviews for future DRI updates. The goal of this evidence scan was not to conduct a formal systematic review; rather, it was to conduct a thorough scan of the peer-reviewed published literature to (1) determine the status of evidence on the nutrient composition and volume of human milk, and (2) identify new evidence on existing DRI nutrients in human milk that could inform future DRI reviews.
APPROACH
The committee approached its task by designing a literature search strategy with the objective of identifying articles on nutrients to provide information for updating the USDA SR Database and to inform future DRI reviews. The committee focused its approach on evidence describing the nutrient composition based on nutrients specified in the DRIs and on the volume of human milk consumed by the infant (milk volume). The DRI framework structure for a systematic review is grounded in the relationship between nutrient intake (exposure) and health outcomes (Russell et al., 2009). However, for this evidence scan, the committee evaluated nutrient composition and milk volume as predictors of nutrient intake for infants broadly; it did not examine specific associations between nutrient intake and infant/child health outcomes (see methodological descriptions below for detailed information). Therefore, for both human milk
nutrient composition and volume, the committee considered criteria to evaluate whether the evidence would be sufficient to inform a decision to undertake a rereview of a nutrient in the DRI process (Brannon et al., 2017). However, this report does not serve as the basis for a specific nutrient review.
Literature Search Criteria
The committee drew from Brannon et al. (2017) as a paradigm to design this evidence scan. To assure the evidence scan was comprehensive, the committee identified subject-matter experts to provide information relative to the development of an analytic framework. The committee convened a 1-day meeting with subject-matter experts in a discussion of issues relevant to the task (see Appendix B for the meeting agenda). The experts were asked to develop their presentations in response to the following objectives for the evidence scan:
- Identify current evidence on the nutrient content and volume of human milk;
- Develop a literature search strategy for the evidence scan; and
- Identify a methodological strategy for assessing the results of the literature search.
Key Questions
After considering information presented by the subject-matter experts, the committee developed the following key research questions to inform a set of inclusion and exclusion criteria, which were subsequently used to assess potentially relevant studies for eligibility.
- What is the nutrient1 composition of human milk produced by healthy2 lactating mothers of healthy,3 singleton, full-term infants up to 12 months postpartum?
- What is the nutrient composition of milk produced by women who are breastfeeding healthy, singleton, full-term infants, up to 5.9 months postpartum?
- What is the nutrient composition of milk produced by women who are breastfeeding healthy, singleton, full-term infants, aged 6 to 12 months?
- What is the volume of human milk consumed by healthy, singleton, full-term, breastfed infants from birth to 12 months of age who are receiving human milk from a healthy mother?
- What is the volume of human milk consumed by healthy, singleton, full-term infants from birth to 5.9 months who are exclusively breastfeeding?
- What is the volume of human milk consumed by healthy, singleton, full-term infants ages 6 to 12 months who are breastfeeding?
___________________
1 Refers to existing DRI nutrients and does not include bioactive substances in foods.
2 The committee decided to accept the authors’ description of the subjects as being “healthy” when this was provided. These considerations meant that some of the reviewed reports could contain data on participants who may not meet the criteria as defined by a DRI committee (such as those who smoke or have obesity).
3 The committee also accepted the authors’ description of the infant as being healthy when this was provided. However, the criteria for adequate growth in infants changed during the 40-year scope of this evidence scan and were poorly described or not at all by the study investigators so it was difficult for the committee to discern if some of the infants studied would be considered to be growing adequately according to current growth standards.
Eligibility Criteria
The committee developed and refined its prespecified criteria for assessing the relevance of identified evidence (see Tables 2-1 and 2-2). The committee used the National Academies’ Research Center as a resource for carrying out the literature search. In its approach, the committee identified a number of relevant articles, which the Research Center staff searched in Medline, Embase, and Scopus (see Appendix C for details on the literature searches). The staff created a list of all the assigned MeSH and Emtree terms, and the committee then reviewed the list of controlled terms and identified additional MeSH terms. The resulting eligibility criteria followed directly from the key questions and are organized using the population, interventions (exposures), comparators, outcomes, and study designs (PI(E)COD) framework in order to focus the key questions into a searchable query. The committee notes that:
- The interventions (I) element of the framework is not relevant, but the exposures (E) element is. As such, nutrient composition (Key Question 1) and milk volume (Key Question 2) are classified as “exposures.”
- The comparators (C) and outcomes (O) elements of the framework are not relevant for this evidence scan.
TABLE 2-1 Prespecified Criteria for Assessing the Relevance of Identified Evidence: Key Question 1: Nutrient Composition of Human Milk
| Component | Criteria |
|---|---|
| Populations |
Human Include
Exclude |
| Exposures |
Milk (e.g., lactation, breastfeeding) containing any of the following:
|
Exclude |
|
| Comparators | Not relevant for this evidence scan |
| Outcomes | Not relevant for this evidence scan |
| Study Designs |
Include
Exclude
|
a Countries are classified according to definitions from the World Bank.
b See https://data.worldbank.org/income-level/high-income (accessed September 28, 2020).
c See https://data.worldbank.org/income-level/upper-middle-income (accessed September 28, 2020).
d The committee acknowledges that contemporary American women have likely received prenatal supplements during gestation and may still have been taking them at the time of milk sampling. Very few reports included information on women’s prior or current use of prenatal or other supplements. As a result, it is possible that the milk composition values reported reflect this exposure and, thus, the results must be interpreted in light of this possibility.
e Galactagogues are pharmaceutical agents, herbal supplements, or foods for which there is limited or no evidence to support breast milk production. Examples include dopamine antagonists such as domperidone or metoclopramide; shatavari, fenugreek, and malunggay; and parsley, ginger, and garlic.
f The committee used lactose rather than carbohydrate as a macronutrient proxy based on the definition in IOM (2002/2005, p. 281).
g The committee chose to exclude data regarding milk produced in the first week postpartum to avoid data that may represent the colostral phase of milk production.
h Examples of bioactive food substances include polyphenols, carotenoids, tocopherols, tocotrienols, organosulfur compounds, and soluble and insoluble fiber.
i The initial literature search was conducted beginning with 1970; however, it was later refined to studies from 1980 onward in order to capture more consistent and up-to-date methodologies in nutrient analyses (see Appendix C).
TABLE 2-2 Prespecified Criteria for Assessing the Relevance of Identified Evidence: Key Question 2: Volume of Human Milk
| Component | Criteria |
|---|---|
| Populations |
Human Include
Exclude
|
| Exposures |
Volume of milk (e.g., lactation, breastfeeding) consumed by infant Exclude
|
| Comparators | Not relevant for this evidence scan |
| Outcomes | Not relevant for this evidence scan |
| Study Designs |
Include
Exclude
|
a See https://data.worldbank.org/income-level/high-income (accessed September 28, 2020).
b See https://data.worldbank.org/income-level/upper-middle-income (accessed September 28, 2020).
c Galactagogues are pharmaceutical agents, herbal supplements, or foods for which there is limited or no evidence to support breast milk production. Examples include dopamine antagonists such as domperidone or metoclopramide; shatavari, fenugreek, and malunggay; and parsley, ginger, and garlic.
d The committee chose to exclude data regarding milk produced in the first week postpartum to avoid data that may represent the colostral phase of milk production.
e The initial literature search was conducted beginning with 1970; however, it was later refined to studies from 1980 onward in order to capture more consistent and up-to-date methodologies in nutrient analyses (see Appendix C).
Search Strategy
Using the prespecified criteria, the committee developed a search strategy that included a search syntax (i.e., specific keywords and controlled vocabulary terms) to be used in searching online databases of the peer-reviewed published literature. The search process is shown in Figure 2-1.
Syntax and Language Terms
The National Academies’ research staff compiled a literature search strategy based on the committee’s prespecified criteria for nutrient composition and milk volume (see Tables 2-1 and 2-2, respectively). Natural language terms were identified by the committee to be used in the search syntax (see Figure 2-1, Step 1). Two separate searches were conducted, one for nutrient composition and the other for milk volume. Only English language reports were searched. The search was carried out in Medline (using Ovid) and included published articles from January 1980 through December 2019. Details of the literature searches are shown in Appendix C.
Internal Assessment
To assess whether the search strategy was either too broad or too narrow, the committee developed an internal check (Step 2). Committee members suggested five to six key articles that they expected to find in the search results. Research Center staff reran the searches and checked whether the search had identified the submitted articles. Additionally, study staff conducted three abridged checks of 200–500 records to determine whether refinements would be needed to the search syntax. Audit results were consistent with expected outcomes; therefore, no further refinements were made to the literature search, and the search syntax for Medline was finalized (Step 3).
Next, the committee translated the search syntax for use in Embase (using Ovid) and Scopus on peer-reviewed literature published from January 1980 through January 2020 (Steps 4 and 5). Concurrently, the searches were rerun in Medline to capture any articles that had been published since the preliminary search in December 2019. Because human milk composition and volume are often studied in tandem, records from the milk composition and volume search sets were identified that addressed both. To eliminate the need to screen these articles twice, they were removed from the search result sets and put into a third search result set. All result sets were uploaded to a systematic review management program (Covidence4) for screening.
___________________
4 See https://www.covidence.org (accessed September 28, 2020).
SCREENING AND DATA ABSTRACTION
Screening
Following the internal assessment, in the first step of the screening process, the Health and Medicine Division (HMD) staff screened the title and abstract of each article retrieved from the literature search using the prespecified criteria presented in Tables 2-1 and 2-2. In the second step, two committee members screened the full text of each article to determine eligibility for data abstraction based on the same prespecified criteria. During the screening process, the committee reevaluated and revised the criteria (see Appendix D) and refined the search strategy. Throughout the screening by committee members, all records were double-screened independently and conflicts were resolved through discussion between the two committee screeners.
Data Abstraction
The data abstraction was carried out using a Web-based (Google) spreadsheet (see Appendix E). At this stage, after the screening process was completed, the committee focused on the methods used to sample and analyze each nutrient in human milk in each study. Studies on milk volume were examined for the methods used to measure milk volume and for infants 0–5.9 months exclusively breastfeeding. Each article was examined by two committee members: one as a primary data extractor and one as a secondary validator who verified extracted data; conflicts were resolved by discussion between the data abstractor and the validator, which confirmed that the study was truly eligible for inclusion. If there was agreement that a study was
eligible, data on the study characteristics, population characteristics, milk sampling methodology, and analysis methodology were abstracted and entered into the spreadsheet.
REFERENCES
Brannon, P. M., C. M. Weaver, C. A. Anderson, S. M. Donovan, S. P. Murphy, and A. L. Yaktine. 2016. Scanning for new evidence to prioritize updates to the Dietary Reference Intakes: Case studies for thiamin and phosphorus. American Journal of Clinical Nutrition 104(5):1366-1377.
Health Canada. 2015. Canadian Nutrient File Standard. https://www.canada.ca/en/healthcanada/services/food-nutrition/healthy-eating/nutrient-data/canadian-nutrient-file-about-us.html (accessed January 8, 2020).
IOM (Institute of Medicine). 2002/2005. Dietary Reference Intakes for energy, carbohydrate, fiber, fat, fatty acids, cholesterol, protein, and amino acids. Washington, DC: The National Academies Press.
Russell, R., M. Chung, E. M. Balk, S. Atkinson, E. L. Giovannucci, S. Ip, A. H. Lichtenstein, S. T. Mayne, G. Raman, A. C. Ross, T. A. Trikalinos, K. P. West, Jr., and J. Lau. 2009. Opportunities and challenges in conducting systematic reviews to support the development of nutrient reference values: Vitamin A as an example. American Journal of Clinical Nutrition 89(3):728-733.
USDA (U.S. Department of Agriculture). 2019. National Nutrient Database for Standard Reference. https://www.nal.usda.gov/fnic/usda-nutrient-data-laboratory (accessed December 13, 2019).
