Key findings

The purpose of this study was to identify common dietary patterns and to describe their sociodemographic factors of adherence in a representative sample from urban sub-Saharan Africa. The most frequently consumed food groups included plant-based staples (rice, pasta, and corn; maize-based foods; vegetables; white bread and cereals); animal-based products (egg, poultry, dairy products, red meat, fish) were rarely consumed. We found three distinct dietary patterns: a meat and egg-based pattern that was strongly correlated with poultry, meaty dishes, red meat, and eggs; a fish-based pattern characterized by fish, palm oil, condiments, and red meat; and a starchy food-based pattern that correlated with rice, pasta, and corn, roots and tubers, maize-based foods, and condiments. Adherence to the meat and egg-based pattern was associated with younger age, male sex, higher education, and an advanced economic situation; the fish-based pattern was associated with female sex, a higher education, and a higher economic status; whereas the starchy food-based pattern was related with younger age, and sharing a house with other adults.

Study population

A noteworthy finding pertaining to the study population is the overrepresentation of women of 64.3% compared to a balanced men to women ratio of 100.6:100 in the total HDSS population [21]. This is a common phenomenon in surveys conducted in the homes of participants. One sensible explanation is that, on average, women might spend more time at home than men [29]. However, the sex ratio in the present study was very similar between the different settlement types, suggesting that this effect had a similar influence, regardless of the neighbourhood.

Dietary habits and the nutrition transition

According to Popkin’s framework of the nutrition transition, shifts in dietary behaviour are associated with economic development, demographic changes and changes in the prevalence of certain diseases or risk factors. In the following section, the results of this study are put into context of the stages in this model, ranging from stage 3 (high consumption of fibre, starchy foods, low consumption of fat and meat) through stage 4 (high in fat, sugar, and processed foods) and up to stage 5 (more health-conscious, consisting of reduced intake of fat and sugar in favour of more fruits and vegetables). [2]

In this study population, absolute intakes of all macronutrients and energy seemed to be much lower in informal settlements as compared to formal settlements. This relative difference between settlement types suggests that the population in informal settlements consumes less food overall. The intake of macronutrients relative to energy intake was similar for the different groups analysed. In the total study population, the median carbohydrate intake of 62 energy% slightly exceed the reference intakes suggested by the European Food Safety Authority of 45–60 energy% [30]. Conversely, total fat intake and protein intake were lower in the present study group than the recommended intakes [30]. In summary, the low intake of protein and fat with high intake of carbohydrates suggests that this population is at an early stage of the nutrition transition (stage 3), holding true not only for the general study population but also for the different settlement types analysed [2].

This interpretation is further corroborated by the predominating foods in our survey. These were starchy staples based on maize and other cereals, and vegetables. At the same time, the consumption of meat, processed meat, added fats, and sweet spreads was rather low.

Finally, the combination of foods consumed, as identified by PCA, and the associated sociodemographic factors of adherence underpin the notion that enhanced economic status leads to modernized dietary practices (referring to higher intake of refined carbohydrates, processed foods, red meat). Certain groups are more inclined to eat a diet richer in animal-based products (the meat- and egg-based dietary pattern) than others. Here, younger people, men, better educated subjects, and those with in a better economic situation have a higher probability of adhering to this diet. Similarly, Table 3 shows that people from formal settlements have higher scores for this dietary pattern. This might suggest that men, better educated people and those in a better economic situation might be the groups that are already on their way to the next stage of the nutrition transition (stage 4 according to Popkin).

Factors of adherence to dietary patterns in sub-Saharan Africa

Compared to other studies from urban sub-Saharan Africa, we found an unusually homogenous meat and egg-based pattern. Nkondjock et al. [31] found a similar pattern made up mostly of bush meat, poultry and red meat in 541 Cameroonian adults. Other studies, however, found meat-heavy dietary patterns that also showed strong association with fish and wheat-based foods [24] or with nuts and legumes [32]. Nkondjock et al.’s meat-based pattern was associated with older age, higher energy intake and BMI, as well as low educational level. This contradicts our finding that the meat- and egg-based pattern was associated with younger age and a high educational level. One reason for this might be that the population studied by Nkondjock et al. (members of the Cameroonian military) overall had a higher socioeconomic status than the population we studied. Whereas in our study population, animal-based products might be hard to access for a large part of the group for economic reasons, this might not have been a limiting factor in Nkondjock et al.’s study. The dietary pattern that included meat and fish in the study by Galbete et al.[24] was associated with living in urban Ghana, as opposed to residing in Europe or rural Ghana, with a higher educational level, male sex, younger age; and more physical activity. This corresponds well to the association we found regarding male sex and younger age being associated with stronger adherence to the meat- and egg-based pattern. Obasohan et al. [32] sought to determine the association of dietary patterns with high blood pressure in Nigerian civil servants but did not identify the meat, nuts, and legume pattern as a significant risk factor.

Mank et al. [33] also found a fish-based dietary pattern in 514 children from rural Burkina Faso. In contrast to the pattern we found, however, it was also strongly associated with maize-based foods, whereas our fish-based pattern had a negative association with maize-based foods.

The starchy food-based pattern in this study corresponds best to the “traditional” patterns found in many studies, most frequently consisting of cereals, grains, roots and tubers, corn, nuts and legumes, fermented foods, and staples, such as tô (the national dish of Burkina Faso consisting of corn, millet or sorghum flour) [10, 13, 24, 34]. One study conducted in 2010 in 330 households in Ouagadougou found one “traditional” dietary pattern through cluster analysis, which was characterized by high intakes of leafy vegetables and local cereals and comprised 71% of the population as opposed to 29% classified in an “urban” dietary pattern (higher intake of fat and sugar). Similar to our findings, the traditional dietary pattern had positive associations with the female sex, lower income, and lower education. [10] Similarly, two studies from Ghana also found associations between “traditional” dietary patterns, lower education, lower income, and rural populations [24, 34]. Comparable combinations of foods to the “traditional” dietary patterns are sometimes also summarized as “health conscious” (in this study characterized by high intakes of foods such as vegetables, fruits, roots and tubers, while being low in snacks and sweetened foods) [35]. Classifying such patterns as health conscious, however, is difficult using this methodology, as important factors to the healthfulness of a diet (e.g., dietary variety) are not considered. This is one of the reasons why we opted to name the dietary patterns we found in a more descriptive manner, mentioning the most highly loading food groups.

Unlike previous literature, we did not identify a “sweets” or “snacking” pattern, characterized by high intakes of sweetened snacks, soft drinks, and similar foods [12, 13, 35, 37]. One reason for this might be the fact that many of the studies identifying this pattern were conducted in children or adolescents and the pattern seems to correlate with younger age [37], whereas our study included only adults starting from the age of 25. In addition, the lack of a snacking pattern in our study supports the conclusion that our study population is still at an early stage of the nutrition transition.

Strengths and limitations of this study

This study stands out through its large sample size of 1,000 participants, drawn from a representative sample within the only urban HDSS in Western Africa, including formal and informal settlements. Using PCA to derive dietary patterns, we did not have to rely on “a-priori” reference values. Furthermore, through this method we were able to take into account the whole diet, not just single nutrients. In contrast to cluster analysis (CA), PCA yields continuous variables as results, suitable for subsequent regression analyses. Unlike CA, PCA does not rely on a reference group, thus not limiting statistical power [38]. Moreover, PCA assigns a score to each participant for every dietary pattern unlike CA which yields uneven, mutually exclusive groups, which might limit interpretability [39]. The detailed, region-specific food-frequency questionnaire allowed to cover a wide range of foods and dishes common in the target population.

Food-frequency questionnaires tend to underestimate energy intake due to the closed list of food items. Therefore, the derivation of absolute nutrient intakes may not be accurate. However, on a population basis and for between-group comparisons, FFQ data can be useful to rank participants according to their nutrient intakes. Energy intake, for example, might be underreported by as much as 20% [40]. Furthermore, the AFPQ has not been validated in a Burkinabe population yet. For this reason, we focused on relative intergroup differences, rather than using absolute intake values. Additionally, food-frequency questionnaires do not take into account the mode of preparation of the foods which might impact nutrient uptake. Further limitations of our study include the subjective decisions involved in principle component analysis, such as deciding the number of components to retain, the rotation method, and setting cut-off values for factor loading scores. This might limit comparability with other studies. While the sample was designed to be representative and sample size was comparatively large, the validity of our results might be impacted by response bias due to some sociodemographic groups being more willing to participate, as demonstrated by the disbalanced sex distribution of the respondents. Also, in retrospective dietary assessment, recall bias can distort the findings, when some groups respond systematically differently than others.

Public health implications

Interventions with the goal of promoting a more healthful diet that can profit from the insights gained in this study range from the individual level to interventions that target the general food environment.

On the individual level, nutrition education interventions have been used successfully to improve nutrition knowledge and physical activity level [41, 42]. Our study suggests that educational interventions in Ouagadougou should focus on promoting moderate intakes of starchy foods, while increasing the intake of protein sources. However, interventions should also highlight the risks associated with high intakes of red meat and promote adoption of alternative protein sources (poultry, pulses and other plant-based protein sources). As the nutrition transition progresses, the population should be encouraged to maintain the current habits of low snacking, and avoiding sweets, alcohol, and red meat. Special attention should be paid to the role of men, younger and more affluent population groups, as those appear to have a stronger inclination for meat- and egg-based dietary patterns. Those groups could be the first to move away from traditional diets and thus, becoming at risk of adopting unhealthy diets.

Interventions that work on the food environment level could improve the availability and desirability of certain foods, developing value chains, accompanied by community work [43]. Alternative sources of protein (other than red meat) should be made available and promoted through marketing campaigns. On the policy level, governments can influence the risk of nutrition-related noncommunicable diseases through interventions such as restricting marketing for unhealthful foods (e.g., sweets and snacks) to prevent those from gaining more popularity in this population), funding research on specific risk factors, and improving monitoring and evaluation of adverse consequences [44]. Taxing highly processed foods can be a means of influencing both obesity, as well as undernutrition [45].

Research Implications

With the aim of understanding determinants of a healthy diet, it is necessary to conduct further research into the interpersonal, environmental and policy levels of influence, as those upstream factors of individual decision making are still poorly understood [46]. With the knowledge of how people eat, the question opens up why they choose to do so and what might be barriers on the interpersonal, environmental, and policy levels that prevent individuals from choosing a more healthful diet. Future research should be dedicated to understanding the health consequences of different dietary patterns. It will be important to understand the relationship between meat-based dietary patterns and cardiovascular diseases or type 2 diabetes. Another important factor that might be target of future research in similar contexts is the role of food processing. Using food processing classification systems, such as the NOVA [47] system to classify the food items present in the dietary patterns found could be helpful in understanding the progress of the nutrition transition.

Our study provides a valid starting point to generate in-depth understanding about the nature of the nutrition transition in Ouagadougou and other urban areas in sub-Saharan Africa. In the future, repeated measurements are desirable to evaluate the changes in dietary behaviour over time [48]. This might allow for new insights into the dynamics of diet and for predictions regarding future trends. Health and demographic surveillance systems, such as the one in Ouagadougou, would be ideal for such studies, offering the infrastructure and representative samples of urban populations.