How dietary intake has been assessed in African countries? A systematic review

ABSTRACT Background: Dietary patterns are often considered as one of the main causes of non-communicable diseases worldwide. It is of utmost importance to study dietary habits in developing countries since this work is scarce. Objective: To summarize the most recent research conducted in this field in African countries, namely the most used methodologies and tools. Methods: A systematic review was conducted on MEDLINE®/PubMed, aiming to identify scientific publications focused on studies of dietary intake of different African populations, in a ten-year period. Papers not written in English/Portuguese/Spanish, studies developed among African people but not developed in African countries, studies aiming to assess a particular nutrient/specific food/food toxin and studies that assessed dietary intake among children were excluded. Findings: Out of 99 included studies, the 24-hour recall and the food-frequency questionnaire were the most used dietary intake assessment tools, used to assess diet at an individual level. It was also observed that often country-unspecific food composition databases are used, and the methodologies employed are poorly validated and standardized. Conclusions: There is an emergent need to improve the existing food databases by updating food data and to develop suitable country-specific databases for those that do not have their own food composition table.


Introduction
Over the last years, we have witnessed a shift in demographics in developing countries, namely in what concerns the lifetime expectancy and the organization of the societies, since there has been a growth in urbanization. This reality has led to changes in people's lifestyles, resulting in a transition from traditional to modern realities, subsequently leading to an epidemiological transition. Developing countries, especially in Africa have shown an increase in the prevalence of Non-Communicable Diseases (NCDs), while Communicable Diseases are still a major challenge, despite the success of vaccination programs (Boutayeb, 2006;World Health Organization -Reginal Office for Africa, 2006;Haregu et al., 2014;Islam et al., 2014). According to the World Health Organization (WHO) Global Status Report (2011), NCDs are responsible for almost 80% of deaths in low and middle-income countries. Although the major cause of deaths in African countries are communicable, maternal, perinatal, and nutritional diseases, NCDs are emerging in an exponential rate, being foreseen a switch of trends a switch of trends in 2030 (World Health Organization, 2011).
Dietary patterns are often considered as one of the main causes of NCDs, so it is of utmost importance to describe the expectable nutritional transition, in order to quantify the impact of diet in this group of diseases. However, in developing countries this work is scarce or insufficiently documented, probably due to logistic and financial constraints. According to Pisa et al. (2014), another reason that justifies the scarcity of this work is the lack of reliable dietary assessment methodologies, which upholds the emergent need for the development, validation and standardization of tools for measuring and monitoring food intake in different countries (Pisa et al., 2014). In this regard some work has been done, namely by the Dietary Exposure (DEX) assessment group (Pisa et al., 2014), which addresses its research to studies on diet and cancer and other NCDs. Its main goal is to develop and to validate dietary methods to assess dietary exposures.
The assessment of dietary intake is imperative to know population's food habits, including the inadequacy prevalence of different nutrients, as well as the study of the relationships between dietary patterns and disease. Dietary assessment may be done at national, household and individual level, when approaching food supply and production, food purchases or food consumption, respectively (Thompson and Byers, 1994;Gibson, 2005). At the individual level, several methodologies may be used, and these may be divided into two major groups: retrospective and prospective methods. Retrospective methods comprise the twenty-four hour Recall (24hR), the Food-Frequency Questionnaire (FFQ), and the Dietary History (DH), while prospective methods include Food/Weighed Records (WR) (Thompson and Byers, 1994;Gibson, 2005). Ideally these tools need to be reproducible and valid in order to assure the consistency and accuracy of measurements (Willet, 1998). The choice of an appropriate method will depend on the aims of the study, the population approached as well as the research team's experience.
This systematic review intends to summarize the most recent research conducted in this field in African countries, specifically in what concerns the most used methodologies and tools.

Methods
The literature search was conducted on MEDLINE Ò /PubMed in order to identify scientific papers focused on studies about dietary intake of different populations, among African countries. This research considers several African countries, from North, East, West, Central and South Africa. In order to narrow down and systematize the search in more recent literature, only original papers published between January 2005 and December 2014 were considered and specific key MeSH terms were used: dietary intake; Africa.
Several papers were identified but not all were considered for the systematic review. The inclusion criteria established were related to: the objective of the study: only studies intending to assess dietary intake; the methodology: only studies with a suitably described methodology; and the language: only papers written in English, Portuguese and Spanish. The exclusion criteria were: studies carried out among African people but not in African countries (for example, African immigrants in other countries); assessment of a particular nutrient or a specific food or food toxin; nonquantitative assessment of the diet; dietary assessment among children; and studies performed at household level. Studies that were focused on micronutrient assessment but further evaluated the contribution of macronutrients were also considered. Figure 1 shows a flowchart representing the paper's selection procedure.
A total of 761 studies were identified in the initial search by using the combination of the key terms mentioned above. Out of these, 221 were excluded by reading the abstract while 72 were reviews that were not included in the present study. After applying exclusion criteria, 68 papers were excluded because they did not involve a quantitative analysis or they did not comply to the established objective, 127 studies were related to a single nutrient or food, 119 investigated only children, 12   (Bunch & Murphy, 1997) and the Kenyan food composition database (Sehmi, 1993) were not performed at an individual level, 29 were carried out outside Africa, 14 were written in other languages than English, Portuguese or Spanish, six did not have a suitable description of the methodology or were carried out in very small and characteristic samples. By checking the reference lists of each of these papers, another six papers, which complied with the inclusion criteria, were identified. Thus, the final number of papers was 99. Tables 1-4 summarize the main methodological issues of the included studies, allowing a comprehensive comparison between them. Papers were divided in the four tables according to the method used for dietary assessment: 24hR (Table 1) FFQ (Table 2), both 24hR, and FFQ (Table 3) and WR (Table 4). In each table, in addition to information about the country where the study was conducted and the year of publication, information about methodological issues, such as the study design, studied sample, dietary assessment methods and particularities are presented. Besides these aspects, the sampling methodology and main limitations of the studies were summarized in order to understand the most difficult challenges that researchers found in the field. Furthermore, the application of innovative technologies on dietary assessment in African countries was explored.

Results
The research retrieved 99 papers (102 studies) carried out among different African populations, namely adults (men and/ or women), adolescents and elderly people. The included papers describe studies from twenty-two different countries, located in different African regions: Algeria (n D 1), Egypt (n D 1), Morocco (n D 2), Sudan (n D 1) and Tunisia (n D 5) (Northern Africa), Ethiopia (n D 6), Kenya (n D 10), Malawi (n D 1), Mozambique (n D 3), Uganda (n D 3), Tanzania (n D 7), Zambia (n D 2) and Zimbabwe (n D 1) (Eastern Africa), Benin (n D 4), Burkina Faso (n D 4), Ghana (n D 4), Mali (n D 1) and Nigeria (n D 2) (Western Africa), Botswana (n D 2), and South Africa (n D 38), (Southern Africa), Cameroon (n D 3) and Democratic Republic of the Congo (n D 1) (Central Africa), each region representing 10% (n D 10), 32% (n D 33), 15% (n D 15), and 39% (n D 40), and 4% (n D 4) of the total sample, respectively. The huge representation of Southern Africa is caused by the high number of studies developed in South Africa, representing 37% (n D 38) of the included papers. This division of African regions is based on United Nation (UN) Statistics Division.

Dietary assessment methods
Almost all of these studies are cross-sectional studies, which capture the dietary practices in a specific population, at a particular point in time (Thompson and Byers, 1994;Gibson, 2005).

Methods for the analysis of food intake data
A large range of software tools for the analysis of dietary data were mentioned in these studies. According to Table 5 it can be observed that there is a preferential selection of food databases of the countries within the same African region. For instance, in Western Africa the Software for Intake Distribution Estimation (C-SIDE) developed by Iowa State University is commonly used, while in Eastern Africa, NutriSurvey is the mainly chosen software. In the Northern African countries, Bilnut Software was used for the majority and in the South, FoodFinder Ò (Grant et al., 1992) was clearly the most utilized software. Several countries had to update these tools with their own typical foods or recipes of composite dishes, when these were not available.
In Western African countries the most used nutritional programs were ESHA Food Processor Ò (Food Processor Diet Analysis and Fitness Software) (Wiig and Smith, 2007;Huybregts et al., 2009;Addo et al., 2011;Pereko et al., 2012) and C-SIDE (Sodjinou et al., 2008;Sodjinou et al., 2009;Becquey and Martin-Prevel, 2010;Zeba et al., 2014). Several authors used other softwares, such as NutriData, developed in California (Olayiwola et al., 2012) et al., 2012), which includes the latest version of the South African Food Composition Database, NDSR (Heimburger et al., 2010;Koethe et al., 2013), and WorldFood Dietary Assessment System (Gewa et al., 2008;Walton et al., 2012) were also used. Softwares such as Programme CANDAT (Powell et al., 2013), Food Meter UK 07 (Waudo et al., 2005) and General Intake Estimation System, developed by The National Food Institute, in Denmark (Hansen et al., 2011) (linked with Composition of Foods Commonly Eaten in East Africa, the UK Nutrient Databank and National Food Composition Tables and The Planning of Satisfactory Diets in Kenya), were also chosen for nutrient analysis.
In Central Africa, the used software tools in Cameroon were Microdiet (Anderson et al., 2011) and Becel Institute Nutrition Software (Dapi et al., 2010) and Lucille food analysis software (Termote et al., 2012) in Democratic Republic of Congo.
Some studies did not mention the use of specific software, only referring the use of food composition databases, as source of information for the nutrient analysis, whose analysis was performed with a tool, such as Microsoft Ò Office Excel or IBM SPSS software for example, to compute dietary data.
Generally in African countries, there is a lack of countryspecific Food Composition Tables (FCTs), and the ones that have their own FCT, do not have it updated. For this reason some countries use FCTs from neighboring countries or use global databases. Examples of cited databases are: USDA Nutrient Database for Standard Reference and others developed by Food and Agriculture Organization (FAO), such as

Sampling methodology
Sampling can be done using different methodologies, depending on the aim of the investigation, on the sample size, among other factors. According to the range of studied papers, random and nonrandom sampling methods were used, random sampling being the most common. Some authors did not describe how the recruitment of subjects was done. Within random sampling the main methods used were stratified sampling, multi-stage sampling and simple random sampling. Non-random convenience sampling was also used.
In studies using a multistage sampling approach, probability proportionate to size method was frequently used in the first selection stage, i.e., in the selection of areas (rural/urban), districts, villages, communities or even quarters. Consequently households were simply randomly selected Ijarotimi and Keshinro, 2008;Sodjinou et al., 2009;Nago et al., 2010;Amare et al., 2012;Olayiwola et al., 2012) or the walk method was used (Alemayehu et al., 2011;Korkalo et al., 2014), for the identification of the subject that fulfilled the inclusion criteria, within each household. Besides these walk methods, the usage of township maps to more easily select residential areas was also utilized as one of the initial methodologies of the sampling procedure (Hattingh et al., 2006;Oldewage-Theron and Kruger, 2011;Kolahdooz et al., 2013;Oldewage-Theron et al., 2014). Few studies (Becquey et al., 2009;Nago et al., 2010;Powell et al., 2013) mentioned the possibility to access residential information in the studied city, which were provided by state agencies, such as the village government, Ministry of Education or the Higher Institute of Population Science. Since it was possible to differentiate segments in populations, in several cases creation of clusters (Waudo et al., 2005;Tessier et al., 2008;Kennedy et al., 2009;Aounallah-Skhiri et al., 2011;Jackson et al., 2012) or stratification of the sample (Kesa and Oldewage-Theron, 2005;Steyn and Nel, 2006;Jackson et al., 2007;Mounir et al., 2007;Vorster et al., 2007;Hogenkamp et al., 2008;Maruapula and Chapman-Novakofski, 2008;Dapi et al., 2010;Anderson et al., 2011;Luke et al., 2011;Steyn et al., 2011;Mbochi et al., 2012;Steyn et al., 2012;Haileslassie et al., 2013;Powell et al., 2013) by environmental or individual factors, such as social strata, income, living in rural or urban area, age or sex, was performed. In some cases, participants were randomly recruited using advertisements which were placed in different and strategic locations, such as church groups, community centers and universities or even in local newspapers (Charlton et al., 2005;O'Keefe et al., 2007;Goedecke et al., 2009). Subjects were recruited in medical clinics, health centers or day care centers; in some papers random sampling was cited ( and in others no sampling method was mentioned (Belgnaoui and Belahsen, 2006;Wiig and Smith, 2007;Oldewage-Theron et al., 2008;Baroudi et al., 2010;Heimburger et al., 2010;Oldewage-Theron et al., 2010;Gibson et al., 2011;Kim et al., 2014).

Discussion
The purpose of this review was to summarize the methodologies and tools used in dietary intake assessment in African countries, in a ten year period, covering steps from the sampling to dietary data analysis.
When we seek to describe the dietary intake of a population, the first required step is to establish a representative sample. Many sampling and subject recruitment methods can be used and these were indeed reflected in the analyzed studies. The major part of the studies applied random recruitment. In the cases in which convenience sampling was performed, various segments of the population were not included and since it is not possible to calculate the total number of excluded people, it is also not possible to determine bias caused by the absence of these individuals in the sample (Gouveia de Oliveira, 2009). Ideally, random methods should be used when the aim is to characterize a population. In the recruitment process, the selected approaches have to be adapted to the population socio-economical and educational capabilities in order to assure adequate response rates and to avoid constraints in participation. The authors of the studies included in this review used some strategies, such as: the description of the study objectives in the population's native language, overcoming language limitations; the possibility to give oral consent for participation in the investigation, overlapping limitations related to high rates of illiteracy; and picking enumerators or volunteers that understood very well the population and their habits aiming to reach their confidence, reducing the possibility of anxiety or suspicion that could be present in such situations (Ngo et al., 2009).

Dietary assessment methods
Implementation of dietary assessment methods may be done in several ways, for instance face-to-face interviews, by telephone, by email, self-administrated or observation when using the weighing method. The selected implementation method is related mainly with social and economic context of the studied areas and the resources available for the research. In this review the majority of the included studies were performed with faceto-face interviews and in two cases, i.e. in South Africa and Cameroon, self-administration was used.
To understand which is the best methodology to choose according to the population and the purpose of the study, and considering that 24hR and FFQ methods rely on respondents' memory, it is important to evaluate the accuracy of memorybased reports. There is cognitive research that confirms that for general people it is easier to describe generic dietary patterns than to describe a specific dietary meal (Wirfalt, 1998). According to Thompson and Byers (1994), in cross-sectional studies generally the most used methodology is the 24hR, as corroborated by this review. Furthermore, as shown in the analysis performed herein, FFQ was the second most used tool. Both have advantages and disadvantages and should be applicable in specific situations. In their review Pisa et al. (2014) also identified the same top chosen dietary assessment tools.
A single 24hR is an indepth-interview that must be administered by trained people. In some of the studies under revision the interviewer was either a nutritionist, a dietitian or a nutrition student who had been previously trained by experts to collect dietary data. Such extensive expert training of the interviewer in state-of-the-science methodology is of extreme importance for obtaining valid and reliable assessments and analyses of dietary intakes. Furthermore, the need for a broad perception and issue awareness is needed to successfully fulfill collection of dietary data using this method which is dependent on the subject's memory. A well-trained interviewer will create the need and relaxed atmosphere for the subject, as well as ask key questions that help subjects remember their intake easily (Willet, 1998). According to Wirfalt (1998), and also Smith and colleagues (Smith, 1991;Smith et al., 1991), more important than closeness of time or number of assessed days are the cues presented to the respondent, which influence aspects of the memory structure that are accessed or activated. There is evidence that the presence of cues prior to method implementation and probes during the assessment, two strategies that were done in the majority of the reviewed interviews, may increase reliability given by individual dietary reports (Smith, 1991;Smith et al., 1991;Thompson and Byers, 1994;Wirfalt, 1998). This was one of the reasons why experts in the area were selected to duly perform the analysis of dietary intake. However, not all studies across different countries could guarantee the collection of data by an expert, probably due to availability of these professionals. Another review, carried out in Africa, also pointed out this limitation (Pisa et al., 2014), referring that in some African countries there is a lack of well-trained nutritionists and dieticians, which makes it challenging, perhaps compromising, the transfer of knowledge and training of interviewers.
The Automated Multiple Pass Method (AMPM), cited by some authors as the reference tool to apply the 24hR, has been tested in different types of populations (Johnson et al., 1996;Jonnalagadda et al., 2000;Conway et al., 2003;Conway et al., 2004) and it has been used in the continuing National Health and Nutrition Examination Survey (NHANES). A single day does not represent the usual consumption because of day-today variation and for that reason several studies conducted multiple recalls. The more recalls are conducted, the greater similitude to usual consumption is obtained and, consequently, better accuracy is achieved. According to Thompson and Byers (1994), the 24hR and the WR,, usually done multiple times, estimate with quantitative accuracy daily food and nutrients intake, while frequency methods, such as FFQ, are limited by their lack of quantitative accuracy. In the reviewed studies, when multiple recalls were applied, they were distributed in non-consecutive days, in order to include week-days and weekend. Ideally all the 7 days of the week should be assessed in order to better represent usual consumption and to avoid possible systematic differences on dietary intake in different days of the week (Willet, 1998). However, the chosen number of days should be considered and decided considering the size of the sample, the purpose of the study, the accuracy desired, the monotony or variety of the diet, as well as the variability of nutrients and foods being assessed (Willet, 1998). Yunsheng and colleagues (Ma et al., 2009) studied how many 24hR are required to describe an individual's intake and they concluded that three is the sufficient number of recalls, since with less than three significant differences in energy estimation were observed and with more than three this parameter did not significantly improve. In some cases there was no possibility of conducting a multiple recall. According to the perception of some authors, the monotony of the diet (Steyn et al., 2011;Nyuar et al., 2012) or the large number of respondents (Kamau-Mbuthia and Elmadfa, 2007) meant that a single recall was enough. In other cases the lack of time and other resources, such as labour and finances were the main causes (Wiig and Smith, 2007;Maruapula and Chapman-Novakofski, 2008).
Validity of the 24hR is usually done by comparison between the reports of the respondents in the recall and the measures recorded or weighed by trained and expert observers. An experiment carried out in Ethiopia (Alemayehu et al., 2011) concluded that in the evaluated setting the 24hR was not an accurate substitute of WR. They concluded that the lack of agreement regarding the number and type of foods between the two methods, caused by memory lapses, and inaccuracies in portion size estimation were the main sources of error. Nevertheless, Gewa et al. (2008) used a to a similar comparison and got different conclusions, supported by higher values of agreement coefficients. In this case, 24hR could be an acceptable alternative to weighing method, however they considered that it was necessary to improve the recall procedure.
Twenty-four hours recall does not cause a huge burden to the respondents as the food records do. Besides, the recall is less likely to modify eating behavior of the respondents, because it is implemented after they have eaten and it does not require literacy, which is necessary to perform a correct, informative and complete food record. In the studied populations this was the major strength of the 24hR and a common reason cited by authors for choosing the recalls rather than the food records. Notwithstanding, when compared with frequency methods these two methods have weaknesses in common, since they are not likely to represent the usual consumption of individuals as reflected in frequency methods.
Therefore, FFQ gives a better idea about the usual consumption because the retrospective period is larger. This period could be since the preceding seven days to the preceding year, for instance. The decision about the ideal time frame is related with two issues, the metabolism of the dietary factor being studied and the physiology/pathophysiology of the outcome (Willet, 1998). If the preceding year is used as time frame the researchers assess the dietary intake throughout the whole year, covering both seasons usually referred as dry season and harvest season. When the reference time frame is shorter, the effect of seasonality is not considered, which was a limitation mentioned by several authors. Nevertheless, some of them mentioned that seasonality probably does not induce major dietary modifications, even though seasonality is believed to have significant effects on the diet and nutritional status. In the context of Africa, especially in rural areas, seasonality is in fact an important issue since the production and consequent consumption of some foods, such as fruit, vegetables and cereals, are directly affected by weather conditions that characterize both dry and wet seasons (Savy et al., 2006;Asombang et al., 2013;Msaki and Hendriks, 2014). These diet modifications can lead to different intake in some nutrients such as vitamins and fat (Faber and Laubscher, 2008;Mitchikpe et al., 2008;Wiesmann et al., 2009;Masibo, 2013). More than half of the studies that used FFQ as dietary assessment tool did not specify FFQ's time frame, although the preceding year was the most cited.
Some of the FFQ utilized by the reviewed authors were specifically created for those studies, hence they included the elaboration of the food list besides other steps. Regarding the food list, the way of organizing food items in a questionnaire determines the answer of the respondent. Wirfalt (1998) cited some studies that had better results regarding reproducibility and validity when food items were organized according to the type of meals they usually consumed rather than when they were organized according to food groups. Most of the reviewed studies that used FFQ organized their dietary information in terms of food groups.
As it is possible to observe from the results section, among the studies developed in other countries besides South Africa, few of them used FFQ as one of the selected tools. In these studies authors had to create a new FFQ because in countries such as Kenya, Mozambique, Uganda, Tanzania, Botswana, Morocco or Ethiopia there is no population-specific FFQ. This shows the need for developing new food frequency tools within the majority of African countries.
Validity of a FFQ is not a very practical and easy process to perform, because it requires a noninvasive observation of total diet of the respondents during a long period, and these validation studies have not yet been done (Thompson and Byers, 1994). What is currently and usually done is the comparison of results from FFQ with that from recalls and records (Thompson and Byers, 1994), a process which for some authors should be called calibration instead of validation (Willet, 1998).
Some of the presented studies were also tested for reproducibility. The previous referred THUSA questionnaire was tested for reproducibility by other authors  in a different population, with Seatswana-speaking adults. They concluded that this questionnaire was reproducible.
Besides these reported cases there is still a lack of validated methods for use in a specific population, and thus the need of updating the validated dietary assessment methods across African countries is emergent. When a validation study is performed, researchers have more confidence in their method since it means that it can actually measure the aspects of diet that it was designed for (Willet, 1998), as long as the study is well-performed.
Estimation of food portion size Estimation of foods portion size is one of the challenging aspects of the recall tools (Thompson and Byers, 1994;Willet, 1998;Venter et al., 2000). In several households within rural settings it is common that all the family eat from a shared bowl, hampering the estimation process (Hudson, 1995;Huybregts et al., 2008;Pisa et al., 2014). There are visual aids which are used to help respondents to accurately report the amounts of food items consumed. In the reviewed studies several tools were used, such as household measures, food models (two-dimension or three-dimension), food photographs and pictures, containers, real food items, among others. Within the studies that used 24hR, some did not mention how this estimation was done leaving less margin to evaluate the associated effectiveness (Mostert et al., 2005;Mounir et al., 2007;O'Keefe et al., 2007;Maruapula and Chapman-Novakofski, 2008;Oldewage-Theron et al., 2008;Heimburger et al., 2010;Gibson et al., 2011;L opez et al., 2012;Pereko et al., 2012;Changamire et al., 2014;May et al., 2014). In fact, there is little data concerning the accuracy of portion size estimation tools. A study of Byrd-Bredbenner and Schwartz (2004) evaluated if using portion size measurement aids (PSMAs) had effect on portion size estimation accuracy, in a group of young adults. The PSMAs were two: one was a life-size card containing pictures of both tennis and golf balls and the other one were both real tennis and golf balls. They concluded that even if the estimation accuracy was improved by the use of PSMAs, estimation errors still remain. In Burkina Faso an album of food photographs was validated for use on food portion size estimation of frequently consumed food items (Huybregts et al., 2008). This validated album, with four photos per one of the eight evaluated food items, could be used in 24hR as a valuable and accurate tool in West African rural settings. Another example of advances in this area is the book of food photographs developed and tested by Venter et al. (2000), with the purpose to be used in the THUSA study. A more recent work (Lombard et al., 2013), also carried out in rural areas of South Africa, focused on the development of a food photography series, mainly geared toward oesophageal cancer patients.
In FFQs, portion size of the food items may either be or not be assessed; in this latter option it can be assumed a common portion size for all subjects. There were few papers among the many reviewed in which authors, having collected the amounts consumed by the respondents, did not report how it was done (Belgnaoui and Belahsen, 2006;Jackson et al., 2007;Vorster et al., 2007;Tessier et al., 2008;Delport et al., 2011;Baroudi et al., 2014;Botha et al., 2014). The most used tools to estimate portion size were household measures, food models (two-dimension or three-dimension) and food photographs, including the validated food photo manual (Venter et al., 2000). Implementation of the FFQ by mail or by telephone was not used in the reviewed studies. Although these possibilities are considered or applied many times in European or American dietary surveys, the socio-economical, political or geophysical conditions found in many African countries may entail natural communication barriers.

Food composition databases
To convert the dietary intake into nutrient intake some components are needed, such as a food composition database, a coding system for matching foods listed with the entries in the food composition database and a software for calculating the nutrients' composition (Thompson and Byers, 1994;Willet, 1998). The right choice of the nutrient database is very important because the estimation of nutrient intake is affected by it. Parameters such as the completeness regarding the included food items and evaluated nutrients are related to the constant updating of the database, so it is imperative to support the analysis on the most recent updated version available (Thompson and Byers, 1994). These nutrient databases are commonly included in computer software programs that process data and calculate individual dietary intake. The choice of the software should be based on the level of specification and detail needed, on the type of foods that are usually consumed by the studied population and on the hardware and software requirements. As mentioned above, and also noted by other reviews (Ngo et al., 2009;Ochola and Masibo, 2014;Pisa et al., 2014), there are few African countries with their own FCT, and countries without their own food table need to use either FCTs from neighboring countries or FAO's FCTs, which decreases reliability of the results. This was one of the most cited limitations by the authors. In this review several softwares were mentioned by the authors, however most of them are composed by the same FCTs, which makes imperative the need of creating updated tools. An example of an effort to improve this lack of countryspecific databases is the study of Becquey et al. (2009), who developed a FCT for Burkina Faso bringing together the information of three sources, namely the FCT for Mali, supplemented by the WorldFood FCT for Senegal and the USDA database. This table was complete for energy, macronutrients and eleven micronutrients. The variability within the same continent is huge, and different lifestyles and typical food patterns are found even within the same country, which makes the finding of uniformity in the FCTs quite challenging, and eventually impossible, and so the countries find themselves obliged to create their own tools. In order to fight against the current lack of updating of these tools it is necessary to join forces geared towards the development of both new and countryspecific FCTs or at least to complete the existing ones.
Besides the limitations that were mentioned along this discussion, limitations related to the adopted methodologies, to self-reporting, to small size sample were also cited. Furthermore, the traditional way of cooking is another challenging question too, because household's women resort to memory and taste rather than follow standard recipes or measurements to cook their dishes, which may hamper a reliable assessment (Wojtusiak et al., 2011).
Concerning improvements in developed countries according to new-technology based dietary assessment methods, it is envisaged that, in the coming years, these innovative tools could be used in African countries. Examples of these methods are a mobile device food record (Zhu et al., 2008) and a system based on images of foods (Schap et al., 2014). Although Wojtusiak and colleagues (Wojtusiak et al., 2011) defend that some methods based on automated analysis of photos, voice recognition and use of simple graphical symbols representing food could be applied in dietary assessment in African countries, there is still a long way to go before that may become a reality. Africa is comprised by a large part of rural areas, some of which even do not have sanitation or electricity and food insecurity is one of the major problems. Africa has a particular social organization characterized by the co-existence of several ethnic groups and societies each one with its own traditions and habits, hampering its conjoint growth and balanced development.

Conclusions and recommendations
Globally, African countries are crossing a challenging public health crisis, which coupled to both weak and poor social and governmental structure leads to major concerns related to health, food security and socio-economic issues. Aiming to counteract the double health burden, characterized by both communicable and non-communicable diseases, a major effort is emerging toward development of health policies and the planning, development and evaluation of nutritional interventional programs.
Data obtained from this review provided a better knowledge of the research works that have been developed in African countries concerning food habits of individuals, strengthening the need to apply a bigger effort in these many nations. As shown in this review, in African countries, there is a lack of periodical, accurate, reliable and country-specific methodologies to assess dietary intake in adults. Major limitations on dietary assessment in Africa were, on the one hand, the deficiency in validated and standardized methodologies to perform the dietary assessment and, on the other, the usage of countryunspecific food composition databases. So, related to the first it is necessary to proceed with validation studies and test for reliability of the used methods, in order to assure the consistency and accuracy of measurements, as well as the confidence therein. Regarding the second cited limitation there is an emergent need to improve the already existing databases by updating food data and to develop suitable country-specific ones for those countries that don't have their own food composition table.
Countries with better social, financial and health resources evidenced more activity in this field and performed more investigations, providing greater data availability. Due to distinct social organization of the continent, with major problems, such as high rates of inadequate education, illiteracy, food insecurity and a frail global health system, the work on this field should be continued and widened to include other African countries. Once surpassed some of these basic challenges it will be desired to follow developed countries' trends in what concerns the usage of innovative tools.