Health effects of commercially- available complementary foods: a systematic review Emma Tzioumis MPH, Melissa Kay MS, MPH, RD, Melecia Wright, Linda Adair, PhD Department of Nutrition Gillings School of Global Public Health University of North Carolina at Chapel Hill 0 Table of Contents Conflict of interest statement ...................................................................................................................... 1 Abstract ........................................................................................................................................................ 2 List of abbreviations ..................................................................................................................................... 3 Part 1 – Breastfeeding replacement ............................................................................................................ 4 1.1 Background ......................................................................................................................................... 4 1.2 Objective ............................................................................................................................................. 2 1.3 Methods .............................................................................................................................................. 2 1.4 Results ................................................................................................................................................. 6 1.5 Conclusion ........................................................................................................................................... 9 Part 2 – Obesity and chronic disease risk .................................................................................................. 20 2.1 Background ....................................................................................................................................... 20 2.2 Objective ........................................................................................................................................... 20 2.3 Methods ............................................................................................................................................ 20 2.4 Results ............................................................................................................................................... 22 2.5 Conclusion ......................................................................................................................................... 27 Part 3 – Nutritional quality ........................................................................................................................ 28 3.1 Background ....................................................................................................................................... 28 3.2 Objective ........................................................................................................................................... 28 3.3 Methods ............................................................................................................................................ 28 3.4 Results ............................................................................................................................................... 30 3.5 Conclusion ......................................................................................................................................... 44 Part 4 – Portion size ................................................................................................................................... 45 4.1 Background ....................................................................................................................................... 45 4.2 Objective ........................................................................................................................................... 45 4.3 Methods ............................................................................................................................................ 45 4.4 Results ............................................................................................................................................... 46 4.5 Conclusion ......................................................................................................................................... 47 Part 5 – Stunting, anemia, micronutrient deficiency risk ......................................................................... 48 5.1 Background ....................................................................................................................................... 48 5.2 Objective ........................................................................................................................................... 48 5.3 Methods ............................................................................................................................................ 48 5.4 Results ............................................................................................................................................... 50 5.5 Conclusion ......................................................................................................................................... 54 Appendices ................................................................................................................................................. 55 1 Appendix 1.1 Search terms for PubMed for question 1 ......................................................................... 55 Appendix 1.2 Table of full texts reviewed and excluded for question 1 ................................................ 57 Appendix 2.1 Search terms for PubMed for question 2 ......................................................................... 61 Appendix 2.2 Table of full texts reviewed and excluded for question 2 ................................................ 62 Appendix 3.1 Search terms for PubMed for question 3 ......................................................................... 63 Appendix 3.2 Table of full texts reviewed and excluded for question 3 ................................................ 64 Appendix 4.1 Search terms for PubMed for question 4 ......................................................................... 66 Appendix 4.2 Table of full texts reviewed and excluded for question 4 ................................................ 67 Appendix 5.1 Search terms for PubMed for question 5 ......................................................................... 68 Appendix 5.2 Table of full texts reviewed and excluded for question 5 ................................................ 69 References .................................................................................................................................................. 72 2 Conflict of interest statement The authors are faculty (LA) or PhD students (ET, MK, MW) in the Department of Nutrition at the Gillings School of Global Public Health. In 2012-14, the Department of Nutrition received an educational grant from Pfizer Nutrition, which supported graduate student stipends for students with a research focus on pediatric nutrition (including the co-authors of this report), and an educational seminar series. In this time frame, Pfizer’s nutrition division was acquired by Nestlé who continued support of the educational program. Under provisions of the agreement with the company, students were free to select and pursue research topics of their choice, and to publish their work without any influence of the company. Funding ended in August 2014, prior to the commissioning of this review. 1 Abstract With the broad goal of informing WHO’s consideration of inappropriate promotion of foods for infants and young children, this systematic review aimed to locate and synthesize results of scientific literature on the health effects of commercially-available complementary foods (CACF) for infants and young children under 2 years of age. The review addresses 5 specific research questions: Q1. To what degree do commercially-available products replace, rather than complement, the intake of breast milk in children 6-23 months of age? Q2. To what degree do commercially-available products consumed by children 6-23 months of age increase the risk of childhood obesity or chronic disease risk factors? Q3. Are commercially-available products nutritionally inferior or superior to home-prepared and/or local foods? Do they contain higher or lower amounts of trans fat, saturated fat, free sugars, or salt? Do they contain higher or lower amounts of essential micronutrients? Do commercially- available products provide nutrients that are generally lacking in the diets of young children? Q4. Are the portion sizes of commercially-available products greater than would be appropriate based on age? Q5. Do commercially-available products reduce the risk of stunting, anemia, or micronutrient deficiencies? We searched the following data bases: PubMed, CINAHL plus, EMBASE, Agricola, CAS (Clinicaltrials.gov), Cochrane, Global Health, WHO Global Library, Business Source Premier, AgEcon, Mintel Oxygen, as well as relevant conference proceedings, and sources of grey literature and reviewed reference lists from highly relevant reviews and articles. Titles and abstracts were extracted and reviewed for relevance. Inter-rater reliability was assessed using Kappa statistics. Outcomes included breastfeeding duration or frequency or breast milk intake for Q1; anthropometric indicators of weight status or biomarkers of chronic disease risk for Q2; nutritional composition of foods or infant intake of nutrients for Q3; portion sizes of complementary foods (CFs) for Q4, and stunting, anemia, and micronutrient deficiencies for Q5. The Population-Intervention-Comparison-Outcome (PICO) for each question was narrowly defined to compare CACFs to home-prepared or locally-available CFs. CACFs were defined as commercial products developed and marketed for infant consumption, or products developed for studies to closely mimic CACFs. We also included investigator-prepared local foods that mimicked home-prepared foods. For feeding trials, these were necessary to allow for blinding, quality control and safety. We found very few studies that met the PICO criteria, most often because they did not report on direct comparisons of CACFs with similar home-prepared foods. For all of the research questions, the evidence was judged to be of low or very low quality. For Q1, we found limited evidence that CACFs did not displace breast milk intake, but their consumption was associated with shorter duration of breastfeeding. For Q2, limited evidence suggests that high protein intake and intake of milk-cereal drinks are associated with higher child BMI. For Q3, results were highly heterogeneous given the wide variety of CACFs and home-prepared foods that were assessed: some CACFs were nutritionally superior to home-prepared or local foods, while the converse was true for others. We found no studies that addressed Q4. For Q5, we found no evidence that CACFs improved infant nutritional status. More evidence is needed to identify the benefits and potential harms of CACFs. Additional evidence might be brought to bear on the research questions by comparing CACFs and home-prepared foods to reference data rather than to each other. 2 List of abbreviations AHRQ Agency for Health Care Research and Quality BF Breastfed, breastfeeding CACF Commercially-available complementary food CF Complementary food GRADE Grading of Recommendations Assessment, Development and Evaluation LMIC Low and middle income countries RCT Randomized controlled trial LNS Lipid Nutrition Supplement 3 PART 1. RESEARCH QUESTION 1 To what degree do commercially-available products replace, rather than complement, the intake of breast milk in children 6-23 months of age? 1.1 Background Owing to the high quality of nutrients and other beneficial aspects of breastfeeding, it is desirable, starting at 6 months, to add high nutritional quality foods to the infant’s diet without displacing breast milk intake. Infants have the ability to self-regulate their intake to maintain energy balance. 1 Thus, concern has been raised that in response to feeding of complementary foods (CF) --- particularly those high in energy density --- infants will consume less breast milk. If the foods displacing breast milk are of lower nutritional quality than breast milk, the infant’s nutritional status may suffer. Evidence on whether breast milk is displaced by CFs should ideally come from experimental studies with accurate and precise quantification of breast milk intake before and after feeding of CFs. Since CFs are recommended for infants 6 months and older,2 it would not be ethical to assign infants to a no-CF food (exclusively breast-fed) comparison group. Thus, relevant studies allowing causal inference are those that examine how different amounts, forms, and composition of CFs provided as part of an experimental protocol influence breast milk intake. No-supplement control groups would be consuming their usual diet, which would need to be carefully quantified. A relatively small number of studies have used this ideal design. The “gold standard” method of quantifying breast milk intake is estimation using deuterium dose to the mother. 3,4 Less accurate methods include test weighing of infants before and after breastfeeding,5 estimation from mother’s report of feeding frequency and duration, and estimation based on usual intakes of infants of given size, age, and growth rate. Respondent burden for test weighing is high, and estimated intakes require researchers to make assumptions that may not be valid for individual infants. In particular, breast milk intake modeled as the difference of estimated total intake and quantitated intake from CFs would not allow investigators to examine how breast milk intake is influenced by CF. Other insights can be obtained by comparing duration of breastfeeding or age-specific prevalence of breastfeeding associated with different amount, type, or composition of CFs. However, no causal inferences can be derived from these types of studies. Such studies would have high potential for endogeneity bias, since mothers make simultaneous decisions about breastfeeding and choice of complementary foods. 1.2 Objective The objective for Q1 is to review evidence on the degree to which commercially-available products replace, rather than complement, the intake of breast milk in children 6-23 months of age. 1.3 Methods 1.3.1 Criteria for considering studies. We did not restrict our search by study design: randomized-controlled trials (RCTs) and observational studies were included. For inclusion, studies had to present evidence on infants and young children, 4 ages 6-23 months, who were still breastfed. The main exposure was consumption of commercially- available complementary food products (CACFs) compared to home-prepared or local CF products. We included studies with a quantifiable aspect of breastfeeding as an outcome variable, including quantification of infant milk intake by one of the methods described in the background, or mother’s reports of duration and/or frequency of breastfeeding. 1.3.2 Search method strategies for identification of studies 1.3.2.1 Electronic searches The following electronic databases were searched to identify relevant studies:  PubMed (2000 – 30 Mar 2015)  CINAHL plus (2000 – 31 Mar 2015)  EMBASE (2000 – 30 Mar 2015)  Agricola (2000 – 31 Mar 2015)  CAS (2000 – 31 Mar 2015)  Clinicaltrials.gov (2000 – 31 Mar 2015)  Cochrane (2000 – 31 Mar 2015)  Global Health (2000 – 31 Mar 2015)  WHO Global Library (2000 – 2 Apr 2015)  Business Source Premier (2000 – 6 Apr 2015) 1.3.2.2 Searching other resources We searched other economics databases (AgEcon, Mintel Oxygen), relevant conference proceedings (Association for Consumer Research, International Health Economics Association), and sources of grey literature (New York Academy of Medicine's Grey Literature Report, Open Grey, PH Partners). We also reviewed reference lists from highly relevant reviews and articles. 1.3.2.3 Search strategy The search strategy was created using a combination of subject headings and free-text keywords to identify publications (see Appendix 1.1). We did not restrict by language or study design. We restricted to studies published after 2000, to ensure that the CACFs analyzed would provide relevant and contemporary evidence. References were extracted and imported into EndNote. Duplicate articles were identified and removed. All articles were assigned a unique record number. 1.3.3 Data collection and analysis 1.3.3.1 Selection of studies Two reviewers independently screened titles and abstracts from electronic databases to identify potentially eligible studies. Prior to starting the work, inter-rater agreement was assessed by screening a subset of titles and abstracts and calculating Cohen's Kappa. Discrepancies were resolved by discussion with a third reviewer. One reviewer screened titles and summaries from the other resources. Full-text articles were obtained, and were independently reviewed by two reviewers according to the inclusion 5 criteria. Discrepancies were resolved with discussion. When the full-text was not available in English it was excluded. 1.3.3.2 Data extraction Excel workbooks created for systematic review data management were utilized for abstract screening and full-text reviews. 6 Record numbers, titles, and abstracts were exported from EndNote into the Excel workbook, and all title and abstract screening and full-text review took place within Excel. Comments on reasons for full-text exclusions were noted in a separate file. Each reviewer had her own Excel workbook, to keep the review process blinded. Two reviewers extracted data into standardized tables. Information was extracted on: general study characteristics (design, location/setting, recruitment strategy), participant characteristics (sample size, age, inclusion/exclusion criteria), description of the commercially-available complementary food and comparison foods, outcome, and results. Two separate risk of bias assessment tables were created (for RCTs vs. observational studies). 1.3.3.3 Assessment of risk bias and quality in included studies Articles included in the review were assessed for quality using AHRQ tools for observational studies, 7,8 and the Cochrane tools for Interventions. 9 The studies included in the final review were highly heterogeneous in design, CFs, and settings, and were not amenable to a meta-analysis of results. Overall quality of the evidence from the final studies was assessed using GRADE evidence profiles. 10 1.4 Results 1.4.1 Literature search We identified 3038 abstracts for initial screening, from which 88 were referred for full-text review. Of these, 8 were included in the final review. Figure 1.1 shows the flowchart of the screening and eligibility evaluation phases. Main reasons for exclusion were that the study did not include CACFs, breastfeeding was not a quantifiable outcome of the study, the comparison group was not infants consuming home- prepared CFs, and the full-text was not available in English (see Appendix 1.2 for a full description of full-texts that were reviewed and excluded from the final set). 6 3038 records identified from all sources 274 duplicates excluded 2765 titles & abstracts reviewed 88 full-text records to be reviewed 2 items not available for review 86 full-text records reviewed 78 Full-text articles excluded: 3 Incorrect population age 9 Did not include CACFs 24 BF outcome not quantifiable 10 Incomplete data 9 Irrelevant 7 Not in English 12 Duplicate 4 Published before 2000 8 publications included Figure 1.1 Flowchart of review process 1.4.2 Characteristics of included studies Of the eight papers included in the review, four were RCTs that assigned alternate CFs to infants and compared their breast milk intake. All were done in LMIC (1 in India, 1 in the Democratic Republic of the Congo (DRC), and 2 in Malawi). Of the four, three used deuterium dose to the mother to quantify breast milk intake, while the fourth relied on mother’s reports of breastfeeding frequency and duration of each feeding episode. Three observational studies, one each in Italy, Germany and Guatemala, related breastfeeding to CFs. An additional paper was a secondary analysis of data from a RCT designed for obesity prevention in Australia. The studies differ markedly in the CFs given to infants and in their comparison groups, as well as methods of quantifying CF intakes. Qualitative summaries of the studies are therefore presented. Detailed descriptions of each included study and their results are included in table 1.1 (RCTs) and table 1.2 (observational studies). Owing to the vast differences in the design and quality of RCTs and observational studies, they are separately evaluated. 1.4.3 Substantive findings from RCTs (Table 1.1) Bhandari et al. 11 provided a milk-cereal supplement produced by a local dairy to Indian infants from 4 to 12 months in a 4-arm study whose primary outcome was infant growth. Investigators also compared the 7