Childhood and adolescence are extremely important developmental stages in life. These early years are when key foundations are laid for adult health and economic well-being(Reference Poulton, Caspi and Milne1,Reference Galobardes, Smith and Lynch2) . The influence of childhood experiences on later adult life is well documented(Reference Poulton, Caspi and Milne1,Reference Ness, Maynard and Frankel3–Reference Maynard, Gunnell and Emmett5) . Therefore, developing healthy nutrition behaviours in childhood may help to prevent not only under-nutrition, stunting and acute child nutrition problems but also chronic, long-term health challenges such as obesity, CVD, type 2 diabetes and stroke(Reference Ness, Maynard and Frankel3,Reference Doak, Visscher and Renders6,Reference Nicklas and Hayes7) . Further, there is increasing evidence of a double burden of malnutrition, characterised by the co-existence of under-nutrition/micronutrient deficiencies along with energy overnutrition or diet-related non-communicable diseases(8). Encouraging healthy nutrition among children and adolescents can be an effective primary prevention strategy for reducing the risk of many non-communicable diseases. Poor child nutrition creates economic and social challenges among the vulnerable(9). Particularly, under-nutrition has been linked to suboptimal brain development, which negatively affects educational performance and economic productivity(Reference Polit10–Reference Hughes and Bryan12). Child malnutrition has been a major health problem in many low- and middle-income countries (LMIC). To reduce global health inequities, the WHO has emphasised the key role of establishing positive early childhood experiences in health and in education(Reference Marmot, Friel and Bell13). Consequently, at the World Education Forum in Dakar, a framework that aimed at Focusing Resources on Effective School Health was launched in recognition of the importance of School Health and Nutrition (SHN) as a priority area for education sector plans(Reference Torres14). Since then, the presence and scope of SHN have grown widely globally(Reference Sarr, Fernandes and Banham15). Specifically, between 2000 and 2015, SHN grew substantially in Education Sector Plans in sub-Saharan Africa (SSA) with school enrolment also rising from 83 % in 2000 to 91 % in 2015(Reference Sarr, Fernandes and Banham15).
Since children spend a substantial proportion of their lives in the school setting, from a public health perspective, it makes sense to make schools as healthy as possible. Schools can offer an optimal setting to promote healthy eating habits(Reference Wang and Stewart16–20). They can provide a unique system for the delivery of cost-effective public health interventions since they have a large reach over the child population(Reference Verjans-Janssen, van de Kolk and Van Kann21). More importantly, in LMIC with limited resources, the evidence indicates that effective school health promotion can offer a strong return on investment(Reference Macnab, Gagnon and Stewart22). Investment in these formative years in childhood can reduce health inequity and create healthy adults. Indeed, schools are an obvious place to facilitate this social investment given the inextricable relationship between education and health(Reference Langford, Bonell and Jones23).
According to the 2015 Millennium Development Goals report, the prevalence of stunting among children has fallen in all regions except SSA, where the numbers increased by about one-third between 1990 and 2013(24). Thus, while in 2015, 24·5 and 15 % of children globally were stunted and underweight, respectively, the African region and South East Asia recorded the highest prevalence of under-nutrition, with the former accounting for 39·4 % of the stunted and 24·9 % of the underweight(24). This clearly indicates that malnutrition remains a major public health concern in the sub-region(8,24–26) . In addition, whereas the average consumption of fruit and vegetables was below the WHO recommendations in all WHO regions, African, South East Asian and South American countries reported the least intake, where schoolchildren typically consumed <300 g/d(Reference Lock, Pomerleau and Causer27). These statistics show that investigating and promoting child nutrition in SSA must be a public health priority, especially if the region is to meet the WHO global nutrition target of improving child nutrition by 2025. As a mediating measure for poor child nutrition in the sub-region, WHO and UN have implemented the Renewed Efforts Against Child Hunger and undernutrition, Scaling Up Nutrition(8) and Accelerating Nutrition Improvement(26) initiatives.
Poor nutrition of schoolchildren can be an important barrier affecting their health status and thus access to education and academic achievements(Reference Sarr, Fernandes and Banham15,Reference Leslie and Jamison17,Reference Gelli, Masset and Folson28) . While the first 1000 d of a child’s life remain crucial, school-aged children have the potential for catch-up growth(Reference Leroy, Ruel and Habicht11) making them a suitable age group to target with well-designed nutrition interventions(Reference Leslie and Jamison17). As a result, a number of initiatives have been promulgated globally to improve SHN, including but not limited to the Focusing Resources on Effective School Health approach(20), Health Promoting Schools(Reference Langford, Bonell and Jones23) and Nutrition Friendly School Initiatives(29). Several studies and reviews have evaluated the effectiveness of SHN programmes globally(Reference Wang and Stewart16,Reference Verjans-Janssen, van de Kolk and Van Kann21,Reference Steyn, Lambert and Parker30–Reference Kristjansson, Petticrew and MacDonald33) . Findings appear to support the effectiveness of multi-component, school-based nutrition interventions (SBNI) in improving nutrition status and nutrition-related Knowledge, Attitude and Behaviour (KAB)(Reference Wang and Stewart16,Reference El Harake, Kharroubi and Hamadeh18,Reference Verjans-Janssen, van de Kolk and Van Kann21) ; however, evidence is inconsistent in terms of the impact of SBNI(Reference Wang and Stewart16,Reference Steyn, Lambert and Parker30,Reference Verstraeten, Roberfroid and Lachat31,Reference De Villiers, Steyn and Draper34) . A preliminary review indicated that no systematic review that evaluates the effectiveness of SBNI in SSA has been published. Therefore, this review aims to evaluate the effectiveness of SBNI involving schoolchildren and adolescents in SSA on child nutrition status and nutrition KAB outcomes.
Methods
The reporting style of this review is based on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analysis) Guidelines(Reference Moher, Liberati and Tetzlaff35).
Eligibility criteria
Inclusion criteria: Studies were included based on:
Setting: Schools in SSA.
Type of interventions: SBNI involving at least one of the following: (a) changes to the school’s physical and social environments; (b) changes to school’s nutrition policies; (c) changes to teaching curricula to incorporate nutrition education and (d) partnership with parents/community.
Participants: Schoolchildren and adolescents, aged 5–19 years.
Study design: Randomised controlled trials (RCT, including cluster RCT), controlled clinical trials, controlled before-and-after studies or quasi-experimental designs with control.
Outcomes of interest: (a) changes in physical indicators/anthropometry; (b) changes in nutrition KAB; (c) biochemical outcomes and (d) psychosocial outcomes. Assessment of measures did not form part of eligibility criteria for initial screening of studies during the electronic search.
Period: No date limit was set. Our last search was done on 20 January 2019.
Language: Studies reported in English.
Publication status: Published.
Exclusion criteria:
Surveys, observational or case studies, theses, policies and commentaries.
Government school feeding interventions which also met the ‘type of study/intervention’ criteria.
Other research reports of included studies.
Information sources
The following electronic bibliographic databases were searched: Cochrane Library, MEDLINE, Embase, PubMed, CINAHL plus, PsycINFO and ProQuest. In addition, Informit, Health Collection and Scopus were searched. We also hand-searched reference lists of previously published systematic reviews on SHN(Reference Wang and Stewart16,Reference Verjans-Janssen, van de Kolk and Van Kann21,Reference Steyn, Lambert and Parker30–Reference Kristjansson, Petticrew and MacDonald33) . Reference lists of included studies and the Public Health Nutrition Journal were also hand-searched. Corresponding authors of included studies, which required further clarification, were contacted through email to find out if the articles we forwarded to them were ‘twin’ reports or if they knew of similar studies to theirs that they could recommend. Authors of six studies(Reference Gelli, Masset and Folson28,Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36–Reference Kugo, Keter and Maiyo39) were contacted, and four of them(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36–Reference Van Stuijvenberg, Dhansay and Lombard38) responded.
Search
All available literature on SBNI in SSA was screened independently by two members of the review team (P.K. and D.E.S.) using study titles and abstracts. The systematic search started in November 2018 and ended on 20 January 2019. Where reviewers were not sure of the eligibility of a study for inclusion, the entire document was downloaded for a full-text screening. The Boolean search terms ‘AND’, ‘OR’ and ‘*’ (for truncation) were applied: ‘school-based nutrition’ or ‘school nutrition intervention*’ or ‘school nutrition program*’ or ‘school meals’ or ‘school breakfast’ or ‘school lunch’ or ‘school diet’ or ‘school food’ or ‘school nutrition education’ AND Sub-Saharan Africa or SSA or Angola, or Benin or…Zimbabwe (all SSA countries were listed, see Supplemental Table 1 in the appendix).
Study selection
Both quantitative and qualitative studies were searched for during the initial search, and no language limit, date limit or design limit was set during the screening stage, although some of those restrictions were applied later using the inclusion criteria. Search strategies were created by a university librarian with expertise in systematic review researching. For transparency and inter-rater reliability, two of the review team members independently screened study titles and abstracts against the inclusion criteria. Full reports for all titles that appeared to meet the inclusion criteria were retrieved. During the electronic database search stage, consensus meetings were held by two of the reviewers (P.K. and D.E.S.) to discuss eligibility of studies about which they had a divergent view. In such cases, a third review team member (J.L.V. or P.L.) was consulted for an opinion.
Data collection process
Data extraction matrix was developed with Microsoft Excel by P.K. and verified by D.E.S. The reliability of the extraction matrix was tested by piloting data entry of the first 10 % of included studies. Data extracted were information on authors’ names, title of study, study aims, participants, intervention, comparators, outcomes, demographics, design, intervention duration and authors’ conclusions. To avoid double counting and to synthesise data from multiple reports on the same intervention (‘companion’ reports), we juxtaposed names of authors, sample size, the outcomes and comparisons used. We considered all reports on a single intervention, but we did not include all companion reports. Only one of such reports was included. The final decision for inclusion or examining the full-text report to determine eligibility was not done by one reviewer but independently by three review members (D.E.S., J.L.V. and P.L.) representing public health physicians, epidemiologists, methodologists and content area experts.
Assessment of risk of bias within studies
To assess studies for risk of bias, we extracted information using the Cochrane ‘Risk of bias’ tool (described in chapter 8 (section 8·5) in the Cochrane Handbook for Systematic Reviews of Interventions(Reference Higgins, Churchill and Chandler40)): that is, random sequence generation, allocation concealment, blinding, incomplete outcome data such as dropouts, and selective outcome reporting. In addition, we assessed study quality using the Effective Public Health Practice Project Quality Assessment Tool for Quantitative Studies(41). Due to the public health nature of the review topic, we only report methodological rigour of studies based on the EPHPP tool (Appendix, Supplemental Table 2). For each included study, each criterion was rated as either ‘strong’, ‘moderate’ or ‘weak’ and then summed up to obtain an overall score (termed as ‘global rating’) for each paper.
Results
Study selection
After initial screening of titles and abstracts, 1041 records were identified through database searching. Sixteen papers were further identified through other sources, such as from reference lists of included studies. The records were exported to EndNotex9 software where duplicates were removed; 602 records remained. After further analysis, 558 records were considered not relevant based on the eligibility criteria and were excluded. Following an assessment of study titles and abstracts, full texts of seventy-six studies were retrieved for further review for eligibility. Out of these, two reviewers (D.E.S. and P.K.) agreed that forty-four studies were potentially eligible for full-text analysis. Subsequently, only fourteen studies(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36–Reference Kugo, Keter and Maiyo39,Reference Van Der Hoeven, Faber and Osei42–Reference Whaley, Sigman and Neumann50) met our pre-specified inclusion criteria; thirty of the potentially eligible articles were excluded as those studies were either ‘companion’ reports of studies already included (n 7); school nutrition surveys/case studies (n 14) which mainly involved assessments of anthropometry and/or nutrition KAB(Reference Fernandes, Folson and Aurino51–Reference Abrahams, De Villiers and Steyn53); analyses of perception and practice of healthy eating among teachers and parents, and development of school food gardens as nutrition tools(Reference Teferi, Atomssa and Mekonnen54–Reference Beery, Adatia and Segantin56); RCT of government school feeding initiatives (n 4)(Reference Gelli, Masset and Folson28,Reference Masset and Gelli57,Reference Sherman and Muehlhoff58) ; or SBNI on pre-schoolers aged <5 years (n 5), these consisted mainly of school-and-community nutrition interventions with parental involvement(Reference Takyi59–Reference Batra, Schlossman and Balan62). Figure 1 presents a flow chart of the review process.
Fig. 1 Flow chart of the review process. SBNI, school-based nutrition interventions; SSA, sub-Saharan Africa; RCT, randomised controlled trials; PICO, participants, intervention, comparators, outcomes
Interventions that were evaluated in our analysis were mainly school-based nutrition programmes focusing on supplementation of school meals with micronutrients and/or assessment of the effects of micronutrients on nutrition status of schoolchildren(Reference Zeba, Martin Prevel and Some36–Reference Van Stuijvenberg, Dhansay and Lombard38,Reference Van Der Hoeven, Faber and Osei42,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) . Others assessed the impact of school nutrition education with/without physical activity programmes on nutrition knowledge, dietary intake patterns and nutrition status(Reference De Villiers, Steyn and Draper34,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47–Reference Lagerkvist, Okello and Adekambi49) . Some papers also focused on promoting healthy dietary choices(Reference Zeba, Martin Prevel and Some36,Reference Van Stuijvenberg, Dhansay and Lombard38,Reference Van Der Hoeven, Faber and Osei42,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Jemmott, Jemmott and O’Leary48,Reference Lagerkvist, Okello and Adekambi49) . None of the included papers specifically provided details on educational, policy or SBNI involving partnership with parents/community in their results. Consequently, we were unable to abstract some of these issues for discussion since they were absent in the original papers.
Study characteristics
Of the fourteen studies, half took place in South Africa and the other seven were from Botswana, Burkina Faso, Kenya, Nigeria and Tanzania. Three were controlled before and after trials(Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Abrams, Mushi and Allen45,Reference Eboh and Boye47) , and 11 (78·6 %) were RCT with three double-blind controlled trials(Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Ash, Tatala and Frongillo46) . All included studies assessed child nutrition status or nutrition KAB, either as primary or secondary outcomes. All but four of the studies(Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47–Reference Lagerkvist, Okello and Adekambi49) reported on anthropometric status. Three studies assessed cognitive outcomes through cognitive tests(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) . Nutrition behaviour outcomes were assessed in seven studies(Reference De Villiers, Steyn and Draper34,Reference Van Der Hoeven, Faber and Osei42,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47–Reference Whaley, Sigman and Neumann50) , and nutrition knowledge was reported by only three(Reference De Villiers, Steyn and Draper34,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47) . Intervention duration varied from 3 weeks(Reference Eboh and Boye47) to 3 years(Reference De Villiers, Steyn and Draper34). All studies were conducted within the last two decades. Specifically, seven were published after the year 2010; only one was conducted before the year 2000(Reference Van Stuijvenberg, Kvalsvig and Faber37). The total number of participants involved in our analysis of this review was 6837 schoolchildren, aged 5–19 years, from 121 schools. In fact, the age group we considered was very wide; however, thirteen out of the fourteen included studies recruited schoolchildren who were aged ≤13 years. Only one study (Jemmot et al.)(Reference Jemmott, Jemmott and O’Leary48) included participants who were older than 13 years. A majority of the participants (4847 out of 6837 pupils, or 71 %) were aged 6–12 years. The implication is that our analysis was mainly on children aged 6–12 years. Consequently, disaggregating results by age (e.g. 5–10 years representing school age and 11–19 years representing adolescents) could not have changed our results significantly, since there was only one study which recruited participants who were older than 13 years. There was a minimum of one school in a study(Reference Van Stuijvenberg, Kvalsvig and Faber37) to a maximum of thirty-nine schools(Reference Zeba, Martin Prevel and Some36). The average number of pupils per study was 488; only two of the studies(Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47) had fewer than 200 participants. See Supplemental Table 3 in the appendix for details on the characteristics of included studies.
Risk of bias within studies
Ten papers (71·4 %) were rated as being of ‘strong’ methodological quality, three (21·4 %) were rated as ‘moderate’, while one (7·2 %) was rated as ‘weak’. Thus, the quality of the evidence of included studies varied, both between studies and across the different domains of potential bias within studies. Many of the studies were rated as being susceptible to ‘high risk of bias’ on the ‘blinding’ criterion. Only six studies stated explicitly that outcome assessors were not aware of the intervention status of participants. Although, as noted above, it was undoubtedly difficult to totally blind outcome assessors and participants since these were public health interventions. Dropout rates varied from 0(Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47) to 43 %. The dropouts were mostly reported in terms of numbers and/or reasons per group in all cases except one(Reference Whaley, Sigman and Neumann50). One study lost nearly 50 % of participants to dropouts(Reference Oosthuizen, Oldewage-Theron and Napier44). Supplemental Table 2 presents details on the ‘risk of bias’ within individual studies.
Intervention effects
Physical outcomes/Anthropometry
Weight and height gains were measured in kilograms and in centimetres, respectively. Study children were classified as stunted, underweight or overweight if their ‘z’ score of height-for-age, weight-for-age and BMI-for-age, respectively, was ±2 sd from the mean of the WHO reference population(63). Of the studies which assessed anthropometric status, eight specifically reported on the prevalence of child stunting(Reference Zeba, Martin Prevel and Some36–Reference Van Stuijvenberg, Dhansay and Lombard38,Reference Van Der Hoeven, Faber and Osei42,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46,Reference Whaley, Sigman and Neumann50) , six reported on wasting(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Van Stuijvenberg, Dhansay and Lombard38,Reference Van Der Hoeven, Faber and Osei42,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45) and seven on BMI(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36,Reference Kugo, Keter and Maiyo39,Reference Van Der Hoeven, Faber and Osei42,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) . Although Kugo et al. (Reference Kugo, Keter and Maiyo39) did not assess prevalence of stunting and wasting, their study and that of Abrams et al. (Reference Abrams, Mushi and Allen45) added mid-upper arm circumference and triceps skin-fold assessment(Reference Kugo, Keter and Maiyo39) to their measures.
In terms of intervention effects on outcomes, Stuijvenberg et al. (Reference Van Stuijvenberg, Kvalsvig and Faber37) presented evidence that fortification of biscuits with Fe, iodine and β-carotene (vitamins) had no favourable effect on anthropometric measures. Other results(Reference Abrams, Mushi and Allen45) indicated that compared with a control beverage, children who consumed fruit-flavoured Beverage Fortified with Micronutrients (BeForMi) of 240 ml servings/week for 8 weeks had significant changes (P = 0·01) in BMI, mid-upper arm circumference, weight-for-age and total weight. In addition, at a follow-up, mean incremental changes in weight (1·79 v. 1·24 kg), height (3·2 v. 2·6 cm) and BMI (0·88 v. 0·53 kg/m2) were also significantly higher in an orange-flavoured BeForMi group than in a non-fortified group(Reference Ash, Tatala and Frongillo46). Moreover, micronutrients fortification or sugar alone in a beverage had a relative lowering effect on weight-for-age relative to controls (micronutrients −0·08; 95 % CI −0·15, −0·01; sugar −0·07; 95 % CI −0·14, −0·002), but when given in combination, the lowering effect was reduced(Reference Taljaard, Covic and van Graan43). Presenting results contrary to the above evidence(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) , analysis from the other studies which assessed anthropometric status observed no significant differences between intervention and control groups on BMI/BMI-for-age(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36,Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) or mid-upper arm circumference(Reference Kugo, Keter and Maiyo39) even after 3 years of school nutrition intervention(Reference De Villiers, Steyn and Draper34).
Nutrition-related Knowledge, Attitude and Behaviour
Nutrition knowledge
All the three studies that reported on nutrition knowledge(Reference De Villiers, Steyn and Draper34,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47) gave evidence in favour of the positive impact of nutrition education on nutrition knowledge. For instance, De Villiers et al. (Reference De Villiers, Steyn and Draper34) found intervention significance (P = 0·02) in an intervention group, both at first and second follow-ups (P = 0·031). In another study aimed at improving dietary intake patterns, correlations linked protein intake to knowledge of proteins, and vitamin C intake to knowledge of fruits and vegetables(Reference Oosthuizen, Oldewage-Theron and Napier44). The nutrition knowledge of the intervention participants improved significantly (P < 0·001) from a total of 45·4 to 58·8 % for all nutrition knowledge questions. Even a long-term measurement still reflected retention of nutrition knowledge, except for topics related to variety in a diet (23·8 %), serving size of specific foods (34·9 %), required daily allowance of specific foods (42·9 %), and fat intake and classification (42·9 %). The results(Reference Oosthuizen, Oldewage-Theron and Napier44) must be interpreted with caution, however, since there was a significant dropout rate (43 %) in this study making the results difficult to generalise. After investigating the effect of school nutrition education programme on nutrition knowledge, Ebo’s study(Reference Eboh and Boye47) corroborated the evidence presented above. They also observed greater increase in nutrition knowledge (P = 0·001) for the intervention group. Regarding the impact of nutrition education on nutrition knowledge, the evidence was gathered from 650 intervention and 620 control participants from twenty-one different schools (Supplemental Table 4 of the appendix).
Nutrition behaviour
Of the seven studies(Reference De Villiers, Steyn and Draper34,Reference Van Der Hoeven, Faber and Osei42,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47–Reference Whaley, Sigman and Neumann50) on this outcome, only two(Reference Eboh and Boye47,Reference Jemmott, Jemmott and O’Leary48) reported a positive impact of SBNI on nutrition behaviour. Improvement in nutrition behaviour such as less sugar intake or more consumption of fruits and vegetables was primary measures in studies that assessed this outcome. Results from one of the nutrition education programmes suggested that nutrition behaviour did not change significantly after 9 weeks of intervention; legumes, fruits and vegetable intake remained low, while refined sugars and fat were still consumed among the intervention group, although mean intake for protein improved significantly(Reference Oosthuizen, Oldewage-Theron and Napier44). Dietary intake analysis by Van der Hoeven et al. (Reference Van Der Hoeven, Faber and Osei42) of the efficacy of green leafy vegetable consumption on micronutrient status also showed no significant differences in energy intake at any of the follow-ups. The median energy intake was 7291 (5768–9960) kJ and 6493 (5258–8457) kJ in the intervention and in the control groups, respectively. Although their ‘HealthKick’ was able to improve nutrition knowledge and self-efficacy significantly, it also had little impact on nutrition behaviour(Reference De Villiers, Steyn and Draper34).
It appears, however, that theory-based and contextually appropriate school health promotion intervention may improve nutrition behaviours. Participants in a cognitive-behavioural nutrition intervention were significantly more likely to have met 5-a-day fruit and vegetable guidelines compared with HIV/STD risk reduction intervention participants in a control group (OR = 1·30, P = 0·008)(Reference Jemmott, Jemmott and O’Leary48). They reported eating approximately 0·54 more servings of fruit (P < 0·05) and 0·77 more servings of vegetables (P < 0·05) than the controls after 12 months. The estimated effect sizes were 0·19 and 0·24 for fruit and vegetables, respectively(Reference Jemmott, Jemmott and O’Leary48). After introducing an orange-fleshed sweet potato meal rich in vitamin A on five occasions for 4 weeks to 3rd and 4th grade intervention participants from twelve schools, their study demonstrated that specific goal setting may help promote nutrition behaviour change(Reference Lagerkvist, Okello and Adekambi49). Thus, directing children to state their intentions to eat a meal could increase the actual proportion of this meal consumed. Besides, the effect on a child’s capability to make changes to their diet (self-efficacy) was found to be significant in a nutrition and physical activity intervention that sought to determine whether nutrition KAB improved after 3 years(Reference De Villiers, Steyn and Draper34). In fact, the study that presented the most favourable results to show evidence of effectiveness of school nutrition education on nutrition behaviour was Ebo’s study(Reference Eboh and Boye47). They found a significant change in compliance in meeting a dietary guideline as well as in meeting food pyramid’s recommendations (P = 0·001). This evidence must also be interpreted with caution, since the methodological rigour of the paper was rated as ‘weak’. So far, the evidence gathered on this outcome point to the conclusion that SBNI do not necessarily influence nutrition behaviour positively.
Biochemical outcomes
The four studies on the BeForMi project included in this review(Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) assessed changes in several micronutrient status indicators, including Hb, Fe, serum retinol (vitamin A1), plasma vitamin B12, riboflavin and serum Zn. Kugo et al. (Reference Kugo, Keter and Maiyo39) tested the efficacy of grounded dried Carica papaya seed mixed with maize porridge on malnutrition and deworming. Their results indicated that a 300 ml maize BeForMi, which contained 10 g of the papaya seed, increased Hb counts of the intervention group (7·14–8·38 mmol/l, P = 0·001). There was also a significant reduction of Ascaris lumbricoides (large round worm) egg count by 63 % (mean 209·7 epg to 75·7 eggs per gram P = 0·002) and Tinea capitis/ringworm infestation (from 54·4 to 34 %, P = 0·002) after 2 months in the intervention group compared with the control that received a one-time 400 mg dosage of albendazole, which is conventionally used for deworming. Evidence from the other food supplementation studies also showed that a BeForMi significantly increased Hb concentration, Fe status indicators (serum ferritin and zinc protoporphyrin) concentrations and vitamin A status(Reference Taljaard, Covic and van Graan43). Using binary logistic regression, controlling for age, sex and baseline Fe deficiency status, they demonstrated that their BeForMi significantly decreased the OR for Fe deficiency (OR 0·20; 95 % CI 0·07, 0·53). The prevalence of Fe deficiency significantly decreased from 29·2 to 5·5 % in children who received the BeForMi(Reference Taljaard, Covic and van Graan43).
Presenting further evidence, Abrams et al. (Reference Abrams, Mushi and Allen45) demonstrated that fruit-flavoured BeForMi with 419 kJ/240 ml blend of twelve micronutrients significantly improved hematologic measures. Fe and vitamin B status were also significantly better, and serum Zn was significantly higher at endpoint in the intervention group. The last BeForMi study also presented similar results. Thus, data from a double-blind placebo efficacy trial of an orange-flavoured BeForMi indicated that among children with anaemia (Hb < 110 g/l) at baseline, there was a significantly larger increase in Hb concentration among participants in the intervention group than those in the control (+9·2 and +0·2 g/l, respectively). In addition, the prevalence of children with vitamin A deficiency dropped from 21·4 to 11·3 % compared with the non-fortified group (20·6–19·7 %)(Reference Abrams, Mushi and Allen45).
Apart from the BeForMi studies, other SBNI have presented similar evidence that food supplementation can improve micronutrient status. To assess the impact of red palm oil (RPO) on vitamin A status, 15 ml RPO was added to school lunch in two test zones. Using HPLC to assess retinol levels, vitamin A status was found to have improved significantly in the RPO group, just as in a positive control that received a single vitamin A capsule of 60 mg (0·77 ± 0·28 to 0·98 ± 0·33 µmol/l). The observed intervention effect was more significant in the RPO group (0·82 ± 0·30 µmol/l to 0·98 ± 0·33 µmol/l) than in a negative control consuming the regular school lunch without RPO (P = 0·001). The efficacy of RPO in addressing vitamin A deficiency was again observed to be more significant in another test zone of the same study, where serum retinol levels increased from 0·77 ± 0·37 µmol/l at baseline to 1·07 ± 0·40 µmol/l one year later (P < 0·001)(Reference Zeba, Martin Prevel and Some36). Further evidence showed that biscuit with RPO as a vitamin A fortificant can also be as effective as biscuit with synthetic β-carotene in improving vitamin A status(Reference Whaley, Sigman and Neumann50). The estimated treatment effect for the synthetic β-carotene biscuit was 2·88 µg/dl (95 % CI 1·75, 4·00) and that of the RPO biscuit was 2·26 µg/dl (95 % CI 1·14, 3·37). A related study also found a significant between-group treatment effect on vitamin A status, Hb, Fe and urinary iodine in favour of participants who received biscuits fortified with micronutrients(Reference Van Stuijvenberg, Kvalsvig and Faber37).
While it appears that green leafy vegetables such as Amaranthus cruentus, Cleome gynandra, Cucurbita maxima and Vigna unguiculate added to a school meal may help address vitamin A deficiency, their effects on other micronutrients have been unclear(Reference Van Der Hoeven, Faber and Osei42). Although a green-leafy vegetable dish contributed 11·6–15·8 mg Fe and 1·4–3·7 mg Zn, no significant intervention effect was found for the dish on micronutrient status. It is important to add that two of the five SBNI among the thirty potentially eligible articles which were excluded in the review process for not meeting the age inclusion criteria(Reference Takyi59,Reference Nawiri, Nyambaka and Murungi60) had however found that intake of dark green, leafy vegetables with fat significantly increased retinol levels (vitamin A) (P < 0·05) among intervention participants(Reference Takyi59), and sun-dried cowpeas with amaranth leaves recipe also enhanced vitamin A status and Hb concentration(Reference Nawiri, Nyambaka and Murungi60) among preschool children. In sum, our analysis of the biochemical indicators as study outcomes showed that all the studies on BeForMi and other SBNI included in this review have presented evidence of the effectiveness of SBNI in improving the micronutrient status of schoolchildren and adolescents.
Psychosocial outcomes
Our analysis also involved assessment of the relationship between SBNI and cognitive performance. Cognitive outcomes comprised: general intelligence (two studies(Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) ), change in arithmetic test scores (three studies(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) ) and verbal comprehension tests involving reading/spelling (three studies(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) ). Findings of Whaley et al. (Reference Whaley, Sigman and Neumann50) showed that supplementation with animal source food plays a key role in the optimal cognitive performance of children. Using the Raven’s Coloured Progressive Matrices to measure general intelligence, they found a ‘most striking’ (sic) significant impact (P = 0·01) of supplementation of a staple diet with meat on general intelligence among Grade One intervention participants, compared with a control after 21 months. Significant group differences were also observed in arithmetic test scores on an adapted version of the Wechsler Intelligence Scales for Children-Revised. However, the effects were neither equivalent across all domains of cognitive functioning nor did different forms of animal source foods produce the same benefits. The study showed no significant difference in verbal meaning test scores(Reference Whaley, Sigman and Neumann50).
Other findings showed that a BeForMi had beneficial effects on cognitive test scores(Reference Taljaard, Covic and van Graan43). A BeForMi improved general intelligence (intervention effect: 0·76; 95 % CI 0·10, 1·42) on the Kaufman Assessment Battery for Children version II test and verbal meaning test scores (1·00; 95 % CI 0·01, 2·00) on an adapted version of the Hopkins Verbal Learning Test. Specifically, there was improvement in planning abilities, number recall, word order, short-term memory recall, story completion and ability to discriminate among words in a familiar setting(Reference Taljaard, Covic and van Graan43). Further evidence confirmed these findings; biscuits fortified with micronutrients (not a BeForMi) resulted in a significant between-group treatment effect in cognitive function tasks such as digit copying, counting letters, reading numbers, counting backwards and verbal fluency(Reference Van Stuijvenberg, Kvalsvig and Faber37). In sum, all three studies of SBNI on cognitive performance found a significant positive impact(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) . Due to differences in data collection methods and measurements, meta-analysis was not feasible in this study.
Discussion
This is the first systematic review of RCT and controlled before-and-after studies to assess the effectiveness of SBNI among schoolchildren and adolescents in SSA. A total of fourteen studies met our inclusion criteria. Duration and complexity of SBNI in SSA over the past two decades have varied. With regard to the impact of SBNI on micronutrient status, studies on beverages fortified with micronutrients (BeForMi)(Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) , and other SBNI on food supplementation/fortification(Reference Zeba, Martin Prevel and Some36–Reference Van Stuijvenberg, Dhansay and Lombard38,Reference Van Der Hoeven, Faber and Osei42) involving 1699 intervention participants, presented evidence of effectiveness of SBNI in improving child micronutrient status. There is sufficient evidence to confirm that food fortification can play a vital role in reducing micronutrient deficiencies. In addition, all studies that assessed cognitive outcomes(Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) , involving a total of 738 and 443 intervention and control participants, respectively, from sixteen different schools, showed effectiveness of SBNI in improving cognitive performance. More specifically, food supplementation with animal source food(Reference Whaley, Sigman and Neumann50) or RPO(Reference Zeba, Martin Prevel and Some36,Reference Van Stuijvenberg, Dhansay and Lombard38) or micronutrients(Reference Taljaard, Covic and van Graan43) significantly improved general intelligence, verbal learning and arithmetic performance of schoolchildren and adolescents. While few nutrition interventions have used comprehensive neuropsychological tests, results from previous systematic reviews corroborate the evidence from our review on assessments of cognitive performance outcome(Reference Hughes and Bryan12,Reference Kristjansson, Petticrew and MacDonald33) . For instance, Kristjansson et al. (Reference Kristjansson, Petticrew and MacDonald33) noted in their review that early micronutrient deficiencies can negatively affect physical, mental and social aspects of child health.
Of the fourteen included studies, only two(Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) observed intervention effect on anthropometry in favour of intervention groups. Thus, although there was evidence to show that SBNI can have a positive impact on anthropometric status(Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) , the majority of studies included in our analysis found no intervention effect(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36,Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) . Specifically, regarding intervention effects on BMI/BMI-for-age, six(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36,Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) out of nine studies which reported on this outcome(Reference De Villiers, Steyn and Draper34,Reference Zeba, Martin Prevel and Some36–Reference Kugo, Keter and Maiyo39,Reference Van Der Hoeven, Faber and Osei42,Reference Taljaard, Covic and van Graan43,Reference Abrams, Mushi and Allen45,Reference Ash, Tatala and Frongillo46) found that SBNI had no significant effect on BMI. This finding is consistent with a previous review of diet interventions on weight status, which found that interventions did not have a significant effect on BMI outcomes(Reference Verstraeten, Roberfroid and Lachat31). On the contrary, of the two studies in the current review that reported an effect on anthropometry, Ash et al. (Reference Ash, Tatala and Frongillo46) observed significant differences between groups for all anthropometric measures; the intervention group gained 0·55 kg more weight, 0·57 cm more height and 0·32 more BMI units. Similarly, Abrams and colleagues(Reference Abrams, Mushi and Allen45) observed significant change in weight, weight-for-age, BMI and mid-upper arm circumference, for the intervention group. Regarding intervention effect on height status/height-for-age, four studies(Reference Zeba, Martin Prevel and Some36,Reference Van Stuijvenberg, Kvalsvig and Faber37,Reference Taljaard, Covic and van Graan43,Reference Whaley, Sigman and Neumann50) reported no intervention effect. The inconclusive evidence of SBNI impact on anthropometry reflects evidence from prior reviews(Reference Verjans-Janssen, van de Kolk and Van Kann21,Reference Langford, Bonell and Jones23,Reference Verstraeten, Roberfroid and Lachat31,Reference Kristjansson, Petticrew and MacDonald33) . In their Cochrane systematic review and meta-analysis, Kristjansson et al. (Reference Kristjansson, Petticrew and MacDonald33) found significant effects of school feeding on weight gain (kg) in lower-income countries but inconsistent results in higher-income countries. For height gain (cm), results from lower-income countries were inconsistent, but in higher-income countries, results were moderate and positive. Further evidence from subgroup analyses indicated that in lower-income countries, height gain was significantly greater for younger children than for older age groups(Reference Kristjansson, Petticrew and MacDonald33). In another meta-analysis on physical activity and nutrition outcomes, the evidence indicated that interventions showed an average reduction in BMI of 0·11 kg/m2, yet the only nutrition study included in the review did not show any intervention effect on BMI(Reference Langford, Bonell and Jones23). As has been noted in previous reviews, the inconsistent evidence of intervention effect on anthropometry may be attributed to baseline malnutrition status or to the short duration of many of the interventions(Reference El Harake, Kharroubi and Hamadeh18,Reference Kristjansson, Petticrew and MacDonald33) . Thus, we might expect to see effects on outcomes such as weight gain even with shorter study durations and on height gain with longer durations(Reference Kristjansson, Petticrew and MacDonald33).
Regarding nutrition behaviour, our analysis suggests that nutrition education may have little impact on nutrition behaviour(Reference De Villiers, Steyn and Draper34,Reference Van Der Hoeven, Faber and Osei42,Reference Oosthuizen, Oldewage-Theron and Napier44) but can improve nutrition knowledge significantly(Reference De Villiers, Steyn and Draper34,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Eboh and Boye47) . Thus, even though there was an improvement in nutrition knowledge, results from other included studies(Reference De Villiers, Steyn and Draper34,Reference Van Der Hoeven, Faber and Osei42,Reference Oosthuizen, Oldewage-Theron and Napier44,Reference Whaley, Sigman and Neumann50) indicated that SBNI could not change dietary intake patterns of participants, and very little variety occurred in diet choices. However, in a health promotion intervention aimed at encouraging health behaviours, the intervention increased fruit and vegetable consumption of adolescents by 1·3 servings/d, compared with the control group(Reference Jemmott, Jemmott and O’Leary48), while specific goal setting also promoted nutrition behaviour change. Similarly, Ebo et al. (Reference Eboh and Boye47) observed that school nutrition education programme improved nutrition behaviour. Our evidence on nutrition behaviour outcome is inconsistent with results of other reviews, which observed significant improvement in nutrition behaviour outcomes(Reference Wang and Stewart16,Reference Steyn, Lambert and Parker30) ; however, it is consistent with results of other meta-analyses which observed moderate improvement in nutrition behaviour(Reference Verjans-Janssen, van de Kolk and Van Kann21,Reference Langford, Bonell and Jones23,Reference Howerton, Bell and Dodd32) . Even if potential gains appear modest due to small effect size, small intervention effects scaled up to large population can produce large public health benefits(Reference Langford, Bonell and Jones23). Evidence from prior studies suggest that although nutrition knowledge may exist, the level of poverty, lack of influence that children have on their food choices(Reference Oosthuizen, Oldewage-Theron and Napier44), food poverty and accessibility could make a complete change to healthier diets somehow difficult(Reference Sherman and Muehlhoff58,Reference Townsend64,Reference Booth65) . In sum, a possible explanation for the inconclusive results regarding intervention effectiveness on nutrition behaviour and anthropometry might be a duration factor as well as the complex nature of eating behaviour, along with limited statistical power(Reference Doak, Visscher and Renders6,Reference Verjans-Janssen, van de Kolk and Van Kann21) . Nutrition behaviour is complex, and it may take time to change dietary habits.
Quality of the evidence
Risks for participants receiving the control intervention or adverse outcomes were generally not reported. In addition, none of the studies explicitly indicated the percentage of relevant confounders which were controlled (either in the design or analysis). As stated earlier, since these were public health and health promotion interventions, total blinding was impossible in many of the studies. In addition, few of the studies discussed existing school nutrition policies or direct parental/community involvement in the development and implementation of the SBNI(Reference De Villiers, Steyn and Draper34,Reference Kugo, Keter and Maiyo39,Reference Van Der Hoeven, Faber and Osei42,Reference Takyi59) . Notwithstanding the above limitations, food supplementation/fortification was generally described as very effective and free of adverse effects. Secondly, all included studies had comparators since they were either controlled before-and-after studies or RCT. Thirdly, all the fourteen studies were rated as ‘strong’ on the ‘data collection method’ criterion on the EPHPP risk of bias assessment tool (Supplemental Table 2 of appendix). This indicated that the data collection tools employed by the primary studies were shown to be both valid and reliable. Moreover, three of the studies(Reference Kugo, Keter and Maiyo39,Reference Taljaard, Covic and van Graan43,Reference Ash, Tatala and Frongillo46) employed double blinding, while the methodological quality of ten of them (70·4 %) was rated as ‘strong’. This makes the risk of bias across these studies low; hence, their evidence can be said to be more reliable.
Implications for health practice, policy and future research
To contextualise these findings, it is important that results from this review be read alongside evaluations of SBNI from other regional contexts which employed different evaluations of study designs other than RCT or controlled before-and-after studies. To address the high incidence of micronutrient deficiencies in LMIC and/or the high incidence of anaemia in SSA in particular(8), the WHO and health professionals may have to intensify food supplementation strategies currently available to schoolchildren in LMIC. Globally, it is important that in countries where schools provide meals to schoolchildren, such meals should be supplemented with vital micronutrients or animal source food to help prevent the double burden of malnutrition. Food and drink fortification with appropriate micronutrients may have double benefits of improving both cognitive performance and nutrition status of schoolchildren. In addition, our findings imply that to effectively design SBNI in the future, policymakers in the education sector planning may need to consider enhancing formal school curricula to include nutrition education since it can positively improve nutrition knowledge.
We recommend that future research should consider investigating the true impact that school nutrition programmes may have on anthropometry and nutrition behaviour, focusing on whether programme intensity and/or duration play any significant role. Specifically, future research must help to find out the impact that nutrition education has on nutrition behaviour since current results on their potential impact are inconclusive. Indeed, the existence of few RCT and controlled before-and-after studies of SBNI in SSA indicates that there might be insufficient evidence from high-quality and analytical school nutrition studies in SSA, and in LMIC in general. This view has also recently (2019) been expressed in a systematic review of food environment research in LMIC(Reference Turner, Kalamatianou and Drewnowski66). This is a challenge, suggesting that there is an urgent need to improve research designs and methods to better understand the effectiveness of public health nutrition programmes in LMIC(Reference Turner, Kalamatianou and Drewnowski66). The implication is that public health researchers and health professionals need to improve the quality of not only school food environment research but also that of the community and national nutrition research. Doing so will undoubtedly be crucial to the design of effective interventions to improve public health nutrition globally.
Limitations
The review process was presented with some methodological challenges. We included only studies published in English, and we also included in our analysis one study with ‘weak’ methodological rigour. Unlike clinical control trials which present more homogenous populations, public health interventions display more heterogeneity. Consequently, the variability among the included studies limited the possibility of meta-analysis on the effect of each factor on child nutrition status. Our reason was that since the factors were measured differently in each study, reporting an estimate for the pool effect would misrepresent the impact of the factors on child nutrition. There was also the possibility of publication bias in the primary studies: studies showing ‘negative’ results are less likely to be written up and submitted, and less likely to be published. Methodological strengths of this review were the use of the PRISMA guidelines(Reference Moher, Liberati and Tetzlaff35) in our reporting, as well as the use of the EPHPP tool(41) to assess the methodological rigour of included studies. This form of assessment has a proven content and construct validity. Our search from more than seven highly recognised electronic databases presents a high level of methodological rigour to the review process. It is also important to note that our review is the first systematic review of RCT and controlled before-and-after studies to assess the effectiveness of SBNI among schoolchildren and adolescents in SSA. Therefore, it provides the best summary to date of the likely average effect of SBNI on nutrition status of schoolchildren and adolescents in the sub-region.
Conclusions
When addressing child malnutrition, evidence from RCT and controlled before-and-after studies of school nutrition interventions in SSA generally confirms the view that the school setting is a very important place to start from. There is strong evidence that supports the positive impact that SBNI can have on cognitive abilities, nutrition knowledge and improved micronutrient status of schoolchildren. There are few existing studies of SBNI in SSA; however, evidence from such studies supports the view that food supplementation is very effective in addressing micronutrient deficiencies in schoolchildren and can improve their overall nutrition status. Secondly, nutrition education may enhance nutrition knowledge, but this may not necessarily translate into healthy nutrition behaviour. This could mean that nutrition knowledge simply has little impact without a facilitating environment. In sum, there is strong evidence to show that SBNI can positively enhance the nutrition status of school-aged children and adolescents. Some evidence also exists to show that SBNI may positively enhance growth and cognitive development. The key conclusion is that there is enough evidence of promise to warrant further trials in these areas.
Acknowledgements
Acknowledgements: None. Financial support: This review received no funding/sponsorship. Conflict of interest: None. Authorship: All authors provided substantial contributions to the development of the manuscript. Conceptualisation: D.E.S. had the original idea for the review; P.K., J.L.V. and P.L. contributed to the overall conception and design. Formal analysis: P.K. Methodology: P.K., J.L.V., P.L., D.E.S. Writing-original draft: P.K. Writing-review and editing: D.E.S., J.L.V. and P.L. provided critical guidance on all aspects of the review and edited the review at all stages. All authors have read and approved the final manuscript. Ethics of human subject participation: Not Applicable.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1368980020000506
