Background

Non-communicable diseases (NCDs) are the leading cause of global disease burden [1, 2], with 80 % of NCD mortality occurring in low- and middle-income countries (LMICs) [3]. Of these, the cardiovascular diseases (CVD), ischemic heart disease and stroke, are the leading causes of death and disability, and are increasing in prevalence [1, 2]. Whereas ischemic heart disease predominates in high-income countries, stroke is the most important CVD in African countries [4]. In 2005, 87 % of stroke death occurred in LMICs, rising to 94 % of stroke deaths in people under 70 years old [5].

Excess sodium intake raises blood pressure, leading to hypertension, the principal preventable risk factor for stroke [6, 7]. Excess sodium is also a major risk factor for other CVDs and for stomach cancer [6–8]. It is possible that populations in sub-Saharan Africa are more vulnerable to the effects of a high sodium diet than other populations due to the greater prevalence of inter-uterine growth restriction, as well as genetic factors [9, 10].

The normal (physiological) requirement for sodium is likely to be between 0.1 and 1.0 g (2.5 g salt) daily [11]. Recommended intake is less than 2.0 g sodium (5.0 g salt) for adults aged 16 and over, and this recommended maximum level of intake should be adjusted downwards in children ages 2-15 based on the energy requirements of children relative to those of adults [12]. The World Health Organization (WHO) Global Action Plan for the Prevention and Control of Non-Communicable Diseases 2013–2020 identifies nine key targets for the reduction of chronic disease, including “a 30 % relative reduction in mean population intake of salt/sodium” [13]. The reason for a focus on a relative reduction rather than aiming to reduce salt intake to recommended levels reflects the observation that most populations have a mean sodium intake that considerably exceeds the 2.0 g recommendation [14, 15].

Given the WHO target, it is important to quantify current population salt intakes. This is particularly important in sub-Saharan Africa where the epidemiological transition is likely to result in dietary changes and a large increase in the prevalence of NCDs. In addition, resources for treatment of salt-associated diseases may not meet population needs so preventative strategies, such as salt reduction strategies, are key to averting the associated morbidity and mortality. To identify populations in which sodium intake is high, and to assess progress against the WHO sodium target, knowledge of sodium intake in sub-Saharan Africa is required.

Systematic reviews examining population salt intake globally have recently been conducted, however, the statistics reported for sub-Saharan African countries were limited [15, 16]. We re-examined this question focusing on sub-Saharan Africa only, in order to maximize the data gathered.

Methods

Search strategy

MEDLINE and Google Scholar were searched on March 2, 2015 using comprehensive search terms [Table 1]. No language limits were applied. While no date limits were applied in the MEDLINE search, the Google Scholar search was restricted to articles published from 1960 onwards. Reference lists of included studies were also searched to identify further studies.

Table 1 Search strategy
Full size table

Inclusion criteria

In order to be considered for inclusion, identified studies must have reported salt or sodium intake based on 24 h or timed urinary collection, or on dietary analysis. Studies reporting spot or overnight urine collection were not eligible for inclusion. Studies were required to include a general population sample or a normotensive population sample from sub-Saharan Africa.

Data extraction and handling

Papers relating to the same study were examined together. Data was extracted independently by two researchers. This was done using a spreadsheet with the following headings:

  1. 1.

    Authors.

  2. 2.

    Year of publication.

  3. 3.

    Year of data collection.

  4. 4.

    Country of study.

  5. 5.

    Details of the sample.

  6. 6.

    Measurement used.

  7. 7.

    Quality assessment: Reliability and accuracy of measurement (after Powles and colleagues, 2013 [15]).

    1. a.

      For urine collection (from high quality to low): U1 – 24 h urine with PABA validation; U2 – 24 h urine with exclusions based on observed/expected creatinine ratio or total urinary creatinine; U3 – 24 h urine with other strict urine collection protocol without use of PABA or creatinine; U4 – 24 h urine with other collection protocol or not recorded; U5 – Less than 24 h urine, but timed in order to correct to 24 h

    2. b.

      For dietary methods: D1 – Multiple short-term diet recalls; D2 – Food Frequency Questionnaire; D3 – Single dietary recall; D4 – Other (stated).

    1. 8.

      Quality assessment – Representativeness of sample population: A – random sampling from explicitly stated sampling frame; B – other sampling strategy.

    2. 9.

      Sample size.

    3. 10.

      Sodium intake (g/day). We extracted data for adults and children, male and female, for urban and rural groups, and for black and non-black ethnicity separately, when these were reported. Where necessary we used the conversion of 1 g sodium = 43.5 mmol sodium, and 1 g sodium = 2.5 g salt. Where both dietary and urinary figures were available we extracted the urinary figures only.

    4. 11.

      Standard deviation of the mean sodium intake.

    5. 12.

      Standard error of the mean sodium intake.

Where there was any discrepancy arising in the data extraction it was resolved by the two reviewers (OO and SO). In cases in which data were not reported in published papers, corresponding authors were contacted to supply these results.

Statistical analysis

Random-effects meta-regression was used to explore the role of sex, year of data collection, country’s economic development, location (urban or rural), and method of measurement (urinary or dietary) as sources of heterogeneity for the estimated salt intakes of adults. For these analyses we treated semi-urban and semi-nomadic populations as rural. Country was used to determine the level of economic development of the study setting based on World Bank definitions at the time of writing this paper [17].

Where year of data collection was not recorded, we assumed three years prior to publication, and where a period covering more than one year was recorded, we used the midpoint for the meta-regression analysis.

Where a standard error of the mean was not reported this was calculated using the standard deviation and number of study participants. In some studies either sample number or standard deviation were missing. If sample number was missing but could be approximated from data included in the paper, this was done. In cases where the range was reported, the standard error of the mean was estimated using the range rule [18].

Sensitivity analyses included excluding studies with the lowest quality measurement and excluding non-black populations (populations in sub-Saharan Africa but not of sub-Saharan African descent). The metareg package in Stata 13 was used for all analyses.

Results

Our search of MEDLINE found 1,057 titles. These were screened and 116 abstracts were retrieved after excluding irrelevant papers. After reading the abstracts, 54 papers were excluded, leaving 62 papers that were retrieved in full. Of these, 37 were found to be relevant. Searching Google Scholar returned 75,700 results. These were ordered by relevance and the first 1,000 titles were reviewed. Three additional papers were retrieved after reviewing these titles. Of these, two were excluded, leaving one paper that included additional reporting of a study previously identified in the MEDLINE search. An additional four papers were found through reference searches. This gave a total of 42 included papers reporting on 34 individual studies [19–60]. Figure 1 is a PRISMA flow chart giving details of the search. Table 2 gives details of included studies.

Fig. 1
figure 1

PRISMA 2009 flow diagram

Full size image
Table 2 Included studies
Full size table

The 34 studies were carried out in 13 countries and published between 1972 and 2013. Dates of data collection were not reported in all studies, but the earliest reported data collection was in 1967 and the most recent was in 2007. Twenty-seven studies examined adults only, four studies examined children only, and three included both adults and children.

Table 3 and Fig. 2 shows sodium intake in adults reported from 30 studies. By extracting data for male and female, for urban and rural groups, and for black and non-black ethnicity separately (when these were reported), this gave 67 population estimates in total. Thirteen of 67 (19.4 %) populations studied had reported sodium intakes below the WHO recommendation. The two lowest sodium intakes were both found in rural Botswana in the 1960s. Of the others, six were found in Kenya in rural populations and recent migrants, and two from an urban and a rural population from Malawi, studied in the 1980s. The remaining three included one female urban population from the Democratic Republic of the Congo, studied in the 1980s; one urban population from the Cameroon; and one rural population from Tanzania and Uganda, studied in the 1990s.

Table 3 Sodium intake (g) in adults
Full size table
Fig. 2
figure 2

Sodium intake (g) in adults (see Table 3). Red line indicates sodium intake of 2 g, the WHO recommended maximum intake. Green markers indicate rural populations. Black markers indicate urban populations. Error bars show standard error of the mean

Full size image

The highest sodium intake was recorded in a Nigerian population in 2006 at over 10 g of sodium per day. It is likely, however, that these outliers are unreliable, as 24 h urine samples were not collected. Instead timed urines were used to calculate possible excretion of a 24 h time frame. Other high sodium intakes (over 5 g/day) were reported in populations in Tanzania, South Africa, and Ghana.

In studies that included estimates for both a rural and an urban population, the urban population estimates were higher in almost every case (Cameroonian estimates in Cooper and colleagues, 1997 [21] were one exception). In studies that included estimates for both male and female populations there were ten in which men consumed more salt than women, six in which women consumed more than men, and one in which their consumption was equal.

Two studies (Kunutsor and Powles, 2009 [23]; Pavan and colleagues, 1997 [56]) were excluded from the meta-regression as they did not report a standard error of mean or enough data to estimate it. Exploring five potential sources of heterogeneity separately suggested there was an association between year of data collection and salt intake; and between location (urban or rural) and salt intake at the 10 % level, with higher salt intakes in more recent studies and higher salt intakes in urban populations than in rural ones (Table 4).

Table 4 Meta-regression entering single covariates
Full size table

Multivariate meta-regression showed a significant association between location and salt intake, with urban populations consuming higher levels of salt than rural populations; and between measurement used and salt intake, with dietary studies finding higher consumption of salt than urinary studies. These associations were robust in sensitivity analyses (Table 5).

Table 5 Meta-regression entering all covariates β coefficient (95 % confidence interval)
Full size table

Table 6 shows sodium intake in children. Urban child populations in Ghana, Nigeria, and the Democratic Republic of Congo were consuming more than the adult recommended intake of salt, as was a population of rural children in Benin. The highest recorded sodium intake was in Nigerian girls aged 12–14 who consumed 3.85 g per day. The lowest sodium intake was found in a male rural population from Zimbabwe who consumed just 0.92 g per day. In each of the three populations for which girls’ and boys’ sodium intake was reported separately, girls had a higher sodium intake than boys.

Table 6 Sodium intake (g) in children Mean (S.D.) n
Full size table

Conclusions

Through extensive systematic searching focusing on sub-Saharan Africa and with no date restrictions, we found 42 papers reporting sodium intakes in sub-Saharan African populations, including seven that examined children. The previous systematic reviews in global populations mentioned in the introduction (Brown and colleagues, 2009 [16]; and Powles and colleagues, 2013 [15]) examined papers published between 1988-2008 and 1980-2011, and identified five and 11 papers reporting sodium intakes in sub-Saharan African populations, respectively. Brown and colleagues [16] did not find any estimates for African children, while Powles and colleagues [15] did not include children in their systematic review.

We have found that sodium intake in many adult populations in sub-Saharan Africa is above the 2 g intake recommended as an upper limit by the WHO, and also above this limit in some populations of children. Indeed, there have been no estimates of sodium intake for adult populations that fell below this 2 g limit reported since the 1990s. Through meta-regression analyses and by doing a within-study comparison, we have found that sodium intake is likely to be higher in urban than in rural populations. This is an important finding because of the trend for urbanization in sub-Saharan Africa, which will put increasing numbers at risk of hypertension and its sequelae, if this association is robust.

Meta-regression also suggested that dietary methods for assessing sodium intake might give higher estimates of sodium consumption than urinary methods in sub-Saharan African settings. Estimates of sodium intake based on dietary and urinary methods of estimation each have specific strengths and weaknesses. Although 24 h urinary excretion is not prone to reporting biases, participant burden is high and this may lead to attrition bias (if quality standards for acceptable collection are stringent) or measurement bias due to incomplete or over-collected urine (where quality standards are lower). In addition, 24 h urinary excretion takes no account of loss of sodium through other means, for example through feces and/or sweat. Timed urine collection allows for a lesser participant burden, but may be biased because of diurnal variation in sodium excretion [61]. Dietary estimates of sodium intake might not be accurate due to recall bias, reporting errors, erroneous food composition tables (for example, because they are not country specific or because they are out-of-date), and/or difficulty in quantifying added salt (including, for example, salt added during cooking but discarded in cooking water, rather than consumed). Our finding that sodium intake estimates are higher based on dietary rather than urinary measures is the opposite of that reported elsewhere [62]. Others have found that estimates based on food diaries, weighed records, food-frequency questionnaires, and 24 h dietary recall underestimated sodium intakes compared with 24 h urine collections. It is worth noting though, that despite the meta-regression results, in each of the three included papers that used both dietary and urinary estimates of sodium intake [22, 32, 46], the estimate from dietary data was lower than from urinary data. It is conceivable that in sub-Saharan Africa more sodium is lost through routes other than urinary excretion, for example, through sweat. This is worth considering given that Powles and colleagues [15] found that African sodium intakes (although above WHO recommendations) were lower than the rest of the world, based on 24 h urinary excretions, uncorrected for non-renal losses.

We did not find a difference between the sexes in sodium intake – in contrast to both global systematic reviews [15, 16] that found men consumed more than women, and boys consumed more than girls. In the child populations identified in our review, girls consumed more sodium than boys, though this finding is based on just two studies with fairly small samples sizes. There is no apparent reason why there would be a different association with sex in sub-Saharan Africa than in the rest of the world. However, the fact that there may be a difference suggests that context-specific research needs to be done to establish whether patterns of sodium intake in sub-Saharan African settings are similar to or different from high-income countries or other LMICs.

In this study we have not examined potassium intake, which may mitigate the effects of high sodium on blood pressure. Understanding sodium intake in sub-Saharan Africa in the context of the whole diet could give additional levers with which to tackle cardiovascular disease risk.

Of 66 estimates of sodium intake for adults identified in this review, only 20 were given a measurement quality score of 1 or 2, and just 28 were rated grade A for sampling strategy. This demonstrates that high quality studies of sodium intake in Africa remain rare. It also means that the implications of our results are limited – it is possible that sodium intake in sub-Saharan African countries differs to the estimates reported here. High quality measurements in representative samples of the general population will be required to identify where and with whom to intervene in order to meet the WHO target of a 30 % reduction in population sodium intake and to demonstrate progress towards this target.