This was a SR and MA study which was conducted to synthesise the evidence and to summarise the extent of existing literature regrading EBF practice, in Northern African countries (Tunisia, Morocco, Algeria, Libya, Egypt, and Sudan).

The principal research question of this SR and MA was: What is known about EBF practice, and what are the principal barriers and explanations of its non-practice in the Northern African countries?

We defined “Exclusive breastfeeding prevalence” as: the proportion of infants less or equal to 6 months of age who are either: -

Exclusively breastfed (infants who had received only breast milk from his/her mother without other liquids or solids except for drops or medicinal syrups, mineral supplements or medicines)

The review was conducted following the Preferred Reporting Items for Systematic Review and Meta Analyses (PRISMA) guidelines. ¹⁸

The PICO question was not fully applicable in our SR because it was not based on interventional and comparative studies. In this SR, the population of interest was the mothers of infants aged 6 months and below. We did not have intervention to compare views; it was a prevalence estimation and not a comparison of two groups. The principal outcome was the frequency (or prevalence) of EBF during the first six months.

Inclusion criteria

Studies were included in the SR and MA if they fulfilled the following eligibility criteria: ‐

Original research studies (primary studies) which have been conducted among mothers in the Northern African countries (Tunisia, Algeria, Morocco, Libya, Egypt, Sudan) on breastfeeding practice or its prevalence including studies of the Knowledge Attitudes Practice type.

Non-Inclusion criteria

Studies were not included if they met the following criteria: ‐

MA or SR studies or comparative and interventional studies

Exclusion criteria

The exclusion criteria were:

Articles with no relevant information about breastfeeding practice in the title or abstract.

Studies conducted at immediate postpartum which evaluate breastfeeding practice at birth without any information about this practice during the first 6 months of infants’ lives.

The different steps and procedure of article selection are represented using the PRISMA flow diagram.

Search strategy and selection criteria

Two reviewers independently searched the published literature until February 2023.We considered any studies that reported the prevalence proportion of EBF among mothers in the NA countries as eligible for inclusion in the analysis. The international databases, including PubMed, Google Scholar, Science Direct, Scopus, and Web of Science were systematically searched. The search strategy was conducted using the following keywords: “Prevalence”, “Exclusive Breastfeeding”, “Breastfeeding”, “Breastfeeding, exclusive”, “Tunisia”, “Morocco”, “Algeria”, “Libya”, “Egypt”, “Sudan”. The search terms were used separately and in combination using Boolean operators including “OR” or “AND”.

Two reviewers screened the abstracts and eliminated any irrelevant articles. The study selection was conducted on two steps; the first was about title/abstract screening and the second was about full-text screening. The two reviewers independently identified all eligible studies, and any disagreements between them were resolved through consensus and after discussion or having the opinion of a third reviewer.

Data was independently extracted from the included studies.

Data extraction

The following information were reported and extracted from the final list of included papers: the authors’ names, the publication year, the study country, the study design, year of data collection, the sample size, the prevalence of EBF in the sample study.

All the identified studies were exported to Zotero citation manager and duplicates were removed.

Measurement of outcome and exposure

The main outcome of interest for this SR was the prevalence of EBF during the first six months of infant’s life (exclusively breastfed for six months or less) which was defined as the frequency of infants 0-6 Months of age who are exclusively breastfed.

The secondary outcome was the prevalence of EBF until the first six months of life, which was defined as the frequency of infants aged ≥ 6 months exclusively breastfed for at least (up to) the first six months of life.

Three principal factors were selected to be analysed as potential factors for EBF practice: ➢

Mother educational level (illiterate or primary vs. secondary or higher)

The Newcastle-Ottawa-Scale (NOS) quality appraisal checklist was used to assess the quality of individual papers. ¹⁹

The NOS assesses the quality of studies based on three main domains: participant selection, comparability of groups, and outcome assessment. Each domain is evaluated according to specific criteria, and points (stars) are assigned based on the quality of each criterion. The total score can then be used to classify studies as ‘good’ if the study has a total score ≥ 7; as ‘fair’ if the study has a total score between [4-6]; or ‘poor’ quality if the study has a total score ≤3 ^{19 – 21} ).

Random Effect Model

The use of a random-effects model is recommended for MA of observational studies. ²² It was used in our MA to calculate the pooled estimate of the global EBF prevalence (the effect size) and the corresponding 95% confidence intervals (CI).

Heterogeneity Measure: I-Squared statistic, Q-statistics and Tau-squared

The heterogeneity of the findings was explored using statistical measures such as Higgins’s I-Squared statistic (I ²), Cochran’s Q-statistics and Tau ² test. Prediction intervals were estimated to provide a range of expected EBF prevalence among NA countries. ²³ The 95% prediction interval using Tau ² test was calculated to assess the weight or the importance of the heterogeneity between included studies and to estimate how the effect size (prevalence) varies across the different included studies. Statistical significance for Cochran’s Q-statistics based on chi-square with a 5% level of significance among reported prevalence was set at P<0.05. ^{24 , 25}

The studies heterogeneity was also explored by leaving out one or more studies from the analyses using the Baujat plot and the influence plot to identify outlier studies, and then comparing results with the main analysis including all studies, ²⁶ to exclude studies that are at high risk of bias. ^{27 , 28}

Forest plots

Forest plots were used to present the MA results showing the effect estimates for each included study and to estimate the global pooled effect (pooled prevalence) with its 95% confidence interval. ²⁹

Subgroup analyses

We performed subgroup analyses, using univariate and multiple meta-regression analyses, to remediate on this heterogeneity and to investigate its possible sources, ²² by countries of included studies, by sample size of included studies (<500; 500-1000; > 1000) and by publication year.

Publication bias

Publication bias across the studies was evaluated using funnel plots, Regression Test and Rank correlation test for funnel plot asymmetry. ^{30 , 31}

The MA was carried out using R-Software (version 4.2.3) ( R Core Team, 2020).

We conducted our search during February 2023. Out of the 2680 articles identified in the systematic search, 2110 remained after the exclusion of duplicates and removing for other reasons. A further 1769 articles were excluded upon screening the title. The next screening step was based on a review of the abstracts. In this stage we only retained 341 articles which were thoroughly assessed for eligibility. The most common reasons for excluding articles were for non-availability of the full text, not conducted in NA countries, not eligible population, incomplete data. After this stage, 16 studies were found to fit our inclusion/exclusion criteria and were selected to be included in the SR and the MA. The flowchart of the literature search according to the PRISMA guidelines is represented in Figure 1.

The main characteristics of the 16 included studies are summarised in Table 1. Included literature was published from six countries (two studies in Tunisia, three in Morocco, two in Algeria, two in Libya, six in Egypt, and one in Sudan). Egypt ranked first (6 studies) and Morocco ranked second (3 studies) in numbers of studies.

Selected studies reported their findings between 1988 and 2017. Most studies used cross-sectional designs.

The sample sizes of the included studies varied between 100 and 1059, with a total sample size of 7849 participants included.

Results of bias assessment performed using the NOS Quality Assessment to critically appraise the literature included within the SR are presented in Table 2.

Of the 16 included studies in the SR, most studies (N=12) were of fair quality, two studies were of good quality and two studies were of poor quality.

Forest plot for pooled prevalence of exclusive breastfeeding practice during the first 6 months

In total, 16 studies were included with an overall sample size of 7849. The overall pooled effect size (prevalence) was 29% with 95% CI [18 %- 42 %] ( Figure 2). The heterogeneity was high (I ² = 99%; 95% CI [99-99] and Tau ² = 1.4; 95% CI [0.7 – 3.5] with p<0.01).

The prediction interval for EBF prevalence ranged from 3% to 85%, with 95% confidence. This prediction interval represents the range of expected EBF prevalence in 95% of settings.

A funnel plot of the estimates is shown in Figure 3. Regression test (p=0.09) and Rank correlation test (p=0.19) for funnel plot asymmetry showed no evidence of publication bias.

To identify outliers studies, Influence and Baujat plots were plotted ( Figure 4 and Figure 5).

Eleven (11) studies were identified as outliers and were removed from the overall analysis. The estimated pooled prevalence for the remaining studies (N=5 studies from Tunisia, Libya, Morocco, and Egypt) was 40.0% (CI 95% [38.0-42.0]), prediction interval [34%- 48%] with homogenous results (Tau ²= 0.0053, I ² = 31%, p=0.22) ( Figure 6).

The pooled prevalence of the remaining 11 studies was 24.0% (CI _95% [12.0; 42.0]), prediction interval [0.01;0.9] with heterogenous results (Tau ²=1.95, I ²=99%, p<0.01) ( Figure 7).

To identify outlier studies among these 11 studies, Influence and Baujat plots were plotted ( Figure 8 and Figure 9).

Seven (7) studies were identified as outliers (among these 11 studies), and after removing them from the overall analysis, the estimated pooled prevalence was 14.0% (CI95% [12.0; 15.0]), prediction interval [0.11;0.17] with homogenous results (Tau ²= 0.0, I2=0%, p=0.91) ( Figure 10).

Table 3 shows the results of the subgroup analysis according to the country of studies and sample size of included studies (<500; 500-1000; >1000). Tunisia had the lowest prevalence of EBF (11%) while Sudan (63%) and Morocco (48%) had the highest prevalence rates ( Table 3).

The forest plots of the 16 included studies by country and by study sample size are represented, respectively, in Figure 11 and Figure 12.

To determine potential sources of heterogeneity and to predict the EBF practice, univariate and multiple meta-regression analysis was conducted by considering the country, sample size, and year of study conduct as covariates, but no factor was identified as statistically significant ( Table 4).

The pooled prevalence of EBF until the first 6 months of life (eight included studies from all 6 countries except Algeria, with an overall sample size of 3564) was 30.0% (CI _95% [14.0; 52.0]), prediction interval [0.01;0.93] with heterogenous results (Tau ²=1.8, I ²=99%, p<0.01) ( Figure 13).

Mother educational level

A total of 10 studies were included in the analysis. The analysis was carried out using the log odds ratio as the outcome measure. The estimated average of observed log odds (OR) ratio based on the random-effects model was equal to 0.42 (95% CI: −0.44 to 1.29) with a prediction interval at 95% of the OR [−2.3-3.1]. Therefore, the average outcome did not differ significantly from zero (z=0.9604, p=0.3369). According to the Q-test, the true outcomes appear to be heterogeneous (p<0.0001) and I ²=94.1%. The Forest and Funnel plots are shown in Figure 14 and Figure 15 respectively.

Maternal employment

A total of 10 studies were included in the analysis. The estimated average of the log odds ratio (OR) based on the random-effects model was equal to 0.25 (95% CI: −0.57 to 1.07) with a prediction interval at 95% of the OR [−2.1-2.6]. Therefore, the average outcome did not differ significantly from zero (z=0.58, p=0.5567). According to the Q-test, the true outcomes appear to be heterogeneous (p<0.0001) and I ²=94.7 %. The Forest and Funnel plots are shown in Figure 16 and Figure 17 respectively.

Delivery mode: (vaginal or cesarean section)

A total of 5 studies were included in the analysis. The estimated average of the log odds ratio (OR) based on the random-effects model was equal to 0.24 (95% CI: −0.08 to 0.57) with a prediction interval at 95% of the OR [−0.28; 0.76]. Therefore, the average outcome did not differ significantly from zero (z=1.4555, p=0.1455). According to the Q-test, there was no significant amount of heterogeneity in the true outcomes (p=0.1760) and I ²=33.2%. The Forest and Funnel plots are shown in Figure 18 and Figure 19 respectively.

Neither the rank correlation nor the regression test indicated any funnel plot asymmetry (p=0.483 and p=0.327, respectively).

To our knowledge, this was the first SR and MA which aimed to assess the global pooled prevalence of EBF englobing the NA countries and to estimate the pooled OR of principal associated factors of its practice.

The results of our SR and MA contributes to the existing body of knowledge and literature on EBF in a context where there is a notable lack of actualised data, of understanding and of implementation of EBF.

Comprehensive search of various scholarly databases, rigorous methodology and analysis methods used like sub-group analysis, univariate and multiple meta-regression analysis were also among the strengths of our SR and MA. Rigorous and advanced analysis to assess and address heterogeneity was done using different ways like choosing the appropriate model to use (Random model), assessing of outliers using influence and Baujat plots and exploring statistical heterogeneity with different tests. Advanced research has been done to produce colorful Forest plots graphics that are not provided by default by the software. Furthermore, our statistical analysis was done with R software one of the most performant available software in terms of meta-analysis methods.

However, this study also had some limitations, which should be properly recognized. The literature review about EBF practice in this region indicated a paucity of primary research; the literature on the topic was limited and not update with old, published data. Moreover, there was a broad variety of studies settings and the overall quality of the methodologies of the included studies was globally not good, especially due to the non-randomization of the included population which suggest sources of selection bias. Methodological differences in studies may have led to a high, statistically significant heterogeneity. However, the limited number of studies focusing on the prevalence and barriers of EBF in the NA region, as well as their heterogeneity, have restricted our ability to generalize the interpretation of the results on the whole region. Consequently, conducting further research studies on EBF using representative sampling methods and more large sample sizes are recommended to have more representative estimates of the prevalence of EBF.

To promote exclusive breastfeeding in Tunisia and other North African countries like Morocco, Egypt, Libya, Algeria and Sudan the following principal recommendations can be considered: 1.

Raise awareness and education: Access to accurate information and education about the benefits of breastfeeding can positively influence breastfeeding practices. Conduct comprehensive public awareness campaigns to educate mothers, families, healthcare providers, and the public about the benefits of EBF. Community-based educational interventions which highlight the short-term and long-term health advantages for both infants and mothers, may have positive impact on increasing breastfeeding initiation and duration among mothers. Dispel common misconceptions and cultural beliefs that may discourage breastfeeding.

By implementing these recommendations, Tunisia and other NA countries can work towards improving EBF rates and ultimately enhance the health and well-being of their populations.

Otherwise, to reduce the unethical marketing of breast-milk substitutes for babies in African countries, several recommendations and perspectives can be proposed ^{60 , 87 , 88} : ➢

Strengthening legal measures: As seen in the report on the International Code of Marketing of Breast-Milk Substitutes, adopting and reinforcing legal measures against inappropriate marketing is crucial. African countries should enact and enforce strict laws that protect the health of mothers and children from misleading marketing practices.

Finally, it is crucial to mention that cultural beliefs and practices play a significant role in shaping breastfeeding patterns. Therefore, further research specific to NA countries would be necessary to identify country specific factors influencing EBF practices in the region. A particular research area should be paid more attention is the conduction of qualitative studies to more in depth understand social factors and cultural beliefs explaining adherence or refusal of this practice. This will help addressing the specificities and particularities of each society having its own traditions, principles and customs and gain a more nuanced understanding of breastfeeding practices in each particular context.