A Systematic Review of the Incidence, Risk Factors and Case Fatality Rates of Invasive Nontyphoidal Salmonella (iNTS) Disease in Africa (1966 to 2014)

This study systematically reviews the literature on the occurrence, incidence and case fatality rate (CFR) of invasive nontyphoidal Salmonella (iNTS) disease in Africa from 1966 to 2014. Data on the burden of iNTS disease in Africa are sparse and generally have not been aggregated, making it difficult to describe the epidemiology that is needed to inform the development and implementation of effective prevention and control policies. This study involved a comprehensive search of PubMed and Embase databases. It documents the geographical spread of iNTS disease over time in Africa, and describes its reported incidence, risk factors and CFR. We found that Nontyphoidal Salmonella (NTS) have been reported as a cause of bacteraemia in 33 out of 54 African countries, spanning the five geographical regions of Africa, and especially in sub-Saharan Africa since 1966. Our review indicates that NTS have been responsible for up to 39% of community acquired blood stream infections in sub-Saharan Africa with an average CFR of 19%. Salmonella Typhimurium and Enteritidis are the major serovars implicated and together have been responsible for 91%% of the cases of iNTS disease, (where serotype was determined), reported in Africa. The study confirms that iNTS disease is more prevalent amongst Human Immunodeficiency Virus (HIV)-infected individuals, infants, and young children with malaria, anaemia and malnutrition. In conclusion, iNTS disease is a substantial cause of community-acquired bacteraemia in Africa. Given the high morbidity and mortality of iNTS disease in Africa, it is important to develop effective prevention and control strategies including vaccination.


Introduction
Nontyphoidal Salmonellae (NTS) are a major cause of food borne infections throughout the developed and developing world [1]. Although infection most often results in self-limited acute gastroenteritis, NTS have been identified as a major cause of invasive bacterial infections in infants and young children in sub-Saharan Africa and HIV-infected individuals of all ages [2,3]. Invasive NTS (iNTS) disease is recognized as a problem in developed countries in young infants, the elderly and immunocompromised [4]. iNTS disease is caused mainly by Salmonella enterica serovars Typhimurium and Enteritidis [5,6].
NTS gastroenteritis is generally understood to be acquired from animal reservoirs, unlike Salmonella Typhi and Salmonella Paratyphi, where the only recognized reservoir is man. Transmission of gastroenteritis-causing NTS to humans can occur by many routes, including consumption of animal food products, especially eggs, poultry, undercooked meat, produce contaminated with animal waste, contact with animals or their environment, and contaminated water [7,8].The African strains responsible for iNTS disease are characterized by genome degradation and appear to be increasingly adapted to an invasive lifestyle [9]. The relative role of animal reservoirs and human to human transmission of strains causing iNTS disease is unclear [10,11]. iNTS disease is diagnosed definitively by blood or bone marrow culture, usually with low sensitivity [12]. It is also impossible to diagnose using clinical symptoms alone due to the lack of pathognomonic features [5,6]. There is currently no commercially-available rapid diagnostic test for iNTS disease.
Bacteraemia is an important cause of severe and often fatal disease globally, especially in developing countries (where it substantially contributes to childhood deaths) [2,13]. Invasive forms of Salmonella disease include enteric fevers (typhoid and paratyphoid fevers) and NTS bacteraemia and are important causes of morbidity and mortality in Asia and Africa [3,14,15]. In a review article of bacteraemia in Africa from 2010, NTS was found to be responsible for 29.1% of all bloodstream infections [16]. Nevertheless, the data on NTS bacteraemia in Africa are limited with no current aggregate data, so it has been difficult to estimate the burden of NTS bacteraemia in Africa. Such information is vital for planning and implementing costeffective solutions [17] to tackle iNTS disease, especially in Africa with its fragile health-care systems. Infectious diseases are a major obstacle to human development in the African region with people suffering from an extensive range of potentially preventable and treatable conditions including invasive Salmonella disease. Adequate evidence on the burden of Salmonella infections would enable health policy makers to make informed decisions on the need for vaccines against Salmonella infections in their country and region.
The aim of the present study is to address this knowledge gap by conducting a review of the available reports on NTS bacteraemia in Africa and to describe the epidemiology of the disease; different authors were resolved by consensus. The full text version either obtained on line or ordered, when only the abstract was available on line, of potentially relevant articles was retrieved and reviewed critically using predetermined inclusion and exclusion criteria for the study. This was done for articles, regardless of the published language. The reference sections of retrieved full text articles were reviewed critically in search of further potential articles for inclusion.
Inclusion criteria. Studies were included if they: 1. Reported NTS isolated by blood culture 2. Were conducted in and recruited subjects from Africa; Exclusion criteria. Studies were excluded if they: 1. Described NTS disease isolated by stool culture only without blood culture result 2. Did not specify serovar of Salmonella enterica isolated (for example, Salmonella Typhi or one of the serotypes causing NTS) Validity assessment. Study validity was established by use of the selection criteria described above, thereby excluding studies that were thought likely to have results that are either inaccurate, not representative of the reported population or could otherwise not be compared with studies included in the analysis. Studies were not excluded on the basis of potential variability of microbiological techniques and identification of culture isolates.

Review of the selected literature
Data extraction and collection process. Relevant descriptive and quantitative variables were extracted from each of the selected article. A standardised template was used for the data extraction in the form of a Microsoft Excel 2013 workbook with each column of the database corresponding to one of the fields in the template. Double data extraction and entry was performed to ensure accuracy. The variables extracted from each article included name of journal, title of article, publication date, study location (including city, country and region in Africa), study period, patient age, and primary eligibility criteria). Quantitative data collected included number of potential study participants, subjects enrolled, participants who had blood cultured, significant pathogens isolated, NTS pathogens isolated; iNTS disease incidence (if stated), CFR of iNTS disease, and proportion of study participants with risk factors such as infection with malaria or human immunodeficiency virus (HIV) (if known). Where publications separately included data derived from more than one cohort of subjects (e.g. different age groups), each unique study was included as a separate entry in the table. The fields used are listed in S1 Table. Data analysis. Data were cleaned and a descriptive analysis performed with the aid of Microsoft Excel 2013. All reported measures of disease frequency were directly drawn from the literature.
Duplicate reports of iNTS disease cases were sorted and removed from the final analysis or, where this was not possible, clearly identified as potentially overlapping (See supplementary S2 Table for listing of relevant reports). The sorting was done by grouping all reports according to country and then comparing variables including study dates, name and location of the study site, age group of subjects reported, number of blood cultures taken and number of NTS isolated across the studies from each country for potential overlaps.
Incidence data were extracted directly from studies where this was stated and summarized according to sub-group classification, subjects, location, study date, groups at risk and possible risk factors. Incidence was derived in a few studies by dividing the total number of NTS isolated by the population at risk per unit time (year) and stated as cases per 100,000 person year of observation or per 100,000 population per year.
Proportion of NTS as the cause of community acquired blood stream infection was calculated using the formula: number of NTS bacteraemic cases divided by total number of significant/pathogenic bacteraemia in the relevant studies. Relevant studies included prospective and retrospective hospital-based blood culture series for subjects presenting with fever with no known focus of infection or groups of subjects not selected for possible associated risk factors as anaemia, HIV, malaria and malnutrition.
CFR data were extracted directly from the studies where given. CFR was calculated using the formula: death amongst NTS bacteraemia cases divided by total iNTS disease cases multiplied by 100. The CFR data were then summarized according to subjects, location, study period and possible risk factors.
Summary data for risk factors associated with iNTS disease in Africa were extracted directly from the studies where available. Statistically-significant measures of association, such as odds ratio (OR), risk ratio (RR) and prevalence proportion were extracted from the studies.

Search results
The online database search performed on PubMed (completed in February, 2015), using the search string and limiting the results from 1 st January 1966 to 31 st December 2014, yielded 16,638 articles. These were retrieved in Quosa and a full text search for the term 'Salmonella' using Quosa yielded a subset of 1,979 articles. The abstract and full text (where available) of the 1,979 articles obtained were reviewed manually for relevance based on our criteria. 177 articles were finally selected from PubMed and entered into the database (Fig 1). We found more publications using our search string with all the countries in Africa included by name, than with "Africa" alone, and by not restricting publications by using 'human' as a search term. The articles retrieved in Quosa were published in 16 languages of which French (216 articles) was the most common following English.
A similar search strategy was used on the Embase database with the initial search results limited to articles in Embase only, humans and the period 2009 to 2014, yielding 8034 articles (in February, 2015). 736 articles were obtained following a full text search for the term 'Salmonella' in Quosa. Thereafter, the full texts of these articles were reviewed manually and 13 additional unique articles not obtained from the initial PubMed search were added to the database (including 9 abstract-only articles).
One hundred and ninety articles [2,3,10,11,13, were obtained following the literature search and used for the descriptive analysis in our review (Fig 1). Following a further review of these articles, 14 (S2 Table) reporting cases or cohorts of iNTS disease already reported in another published article were excluded from the quantitative analysis, leaving 176 articles including 12 studies with possibly overlapping data (S2 Table). These 176 studies included 223 distinct subject cohorts.
The reports varied in their methodology. Three of the 176 articles were reports from ill subjects in a longitudinal, community-based surveillance. 159/176 (90%) of all articles retrieved are reports of isolates obtained from studies conducted on ill subjects presenting to a health facility setting (hospital or clinic). In fourteen of the reports, we could not determine the original basis for selecting the subjects (E.g., retrospective analysis of microbiology laboratory samples, follow up analyses of bacteraemia in patients previously selected for another study). However, most of these 14 were probably from ill subjects presenting to a health facility.
Only 9 (5%) were designed to derive an estimate of the population-based incidence of iNTS disease in Africa; 76 (43.2%) were prospective hospital-based studies with patients recruited during the course of the study; 22 (12.5%) were retrospective studies with analysis of existing hospital or laboratory records; and the remaining 69 (39.2%) are case reports, series, conference abstracts or outbreak reports.
Geographical location of sites reporting iNTS disease in Africa iNTS disease was reported in 33 countries across the five geographical regions of Africa based on the United Nations classification (Fig 2). The majority, 53% (94/176) of the reports were from the Eastern Africa region (including 28 reports from Kenya, 24 from Malawi, 16 from Tanzania and 10 from Uganda); 26% (46/176) from the Western Africa region (including 9 reports from Cote d'Ivoire, 9 from Nigeria and 8 from Ghana); 10% (17/176) from Central Africa (including 11 reports from Democratic Republic of Congo, 3 from Gabon and 2 from Central African Republic); 7% (12/176) from Northern Africa (including 5 reports from Tunisia and 2 each from Morocco and Algeria); and 4% (7/176)from Southern Africa (South Africa) (S3 Table).
Details of the study population or study site setting could be extracted from only 113 of the 176 reports obtained. 50.4% (57/113) of the reports were from urban sites, 38.9% (44/113) from rural sites and 10.6% (12/113) from both urban and rural sites (S4 Table).

Reports based on year of publication
The earliest report of iNTS disease was published in 1966 [69] and the numbers of reports of iNTS disease increased with time to a peak of 18 reports in 2011 (Fig 4)

Incidence of iNTS disease in Africa
Fourteen reports on the incidence of iNTS disease in Africa were obtained from eight countriesthe Gambia, Ghana, Kenya, Malawi, Mozambique, South Africa, Tanzania & Uganda, spanning three regions, Eastern (ten studies), Southern (one) and Western (three) ( Table 1). The incidence data were obtained using different methods of estimation and were from different population groups including specific age groups, HIV-infected patients and subjects with sickle cell disease. Overall, the estimated incidence of iNTS disease ranged from 1.4 per 100,000 population/year (in South African individuals of all ages in 2003 to 2004) to 2,520 per 100,000 population per year in children < 5 years of age from Ashanti, rural Ghana, in 2007 to 2009. The incidence rates were much higher in Eastern and Western Africa compared to Southern Africa. Across the studies, the estimated incidence rates were higher amongst HIVinfected subjects [123,131], subjects with sickle cell disease [174], young children [122,160,164] and in a rural setting compared to an urban setting [164].
Three studies estimated iNTS disease in the community. These studies were conducted on subjects from all age groups, and without bias for risk factors such as HIV infection, malaria and anaemia. Estimates obtained were 1.4, 164 and >600 per 100,000 population per year from South Africa, Malawi and Ghana respectively [11,83,184].

Proportion of NTS as a cause of community acquired blood stream infection in Africa
Fifty six studies, describing a total of 114,634 blood cultures (S5 Table) from four regions were eligible for analysis: hospital-based studies from Africa investigating the organisms causing bacteraemia in the community and not in any specific risk group. Eligible studies were from four of the five regions (all except the Northern region) and sixteen countries in Africa. NTS proportions varied between and within regions. Regional averages ranged from 8% in Southern Africa to 38% in Central Africa with an overall average of 25%. In Eastern Africa, with an average of 27%, the range was 9% in Ethiopia to 39% in Malawi while in Western Africa, with an average of 18%, the range was 8% in Nigeria to 34% in Burkina Faso.
Across countries, NTS was a cause of about 8% (lowest value) of community acquired bacteraemia in both Nigeria and South Africa to a peak value of 45% in one study from Central African Republic involving 131 blood cultures (Fig 5). Variations in NTS proportion within country were found in Malawi, Kenya, Tanzania and Uganda. This suggests that it might be difficult to directly extrapolate the burden of iNTS disease within a country as well as from one country to another.
An analysis of the relative frequencies of different organisms causing community acquired bacteraemia in Africa was performed and it showed that NTS, Staphylococcus aureus and Streptococcus pneumoniae are most prevalent organisms isolated over time in Africa (Fig 6). Since the selection of articles included in this review excluded blood culture series without NTS,  there is an element of bias favouring NTS compared to other organisms. However, the review of Reddy et al [16] also indicated the importance of iNTS as a cause of community acquired blood stream infection amongst adults and children in Africa [16].

Risk factors associated with iNTS disease in Africa
The risk factors known to be associated with iNTS disease in Africa include HIV infection, malnutrition, malaria, young age, anaemia and rural setting ( Table 2). Seven studies from Africa showed a positive association between HIV and iNTS disease in Africa. The estimated odds ratio (OR) from the studies for HIV-infected individuals developing iNTS disease compared with HIV-uninfected individuals ranged from 3.2 to 48.2. The strong association between HIV infection and iNTS disease in Africa was described in 1990 when Gilks et al reported an OR of 48.2 (confidence interval  in Nairobi, Kenya [81]. Malnutrition had a positive association with iNTS disease in five studies (Table 2), with the earliest study published in 2005. The OR for iNTS disease occurring in children with malnutrition compared with children without malnutrition ranged from 1.44 to 2.42. Plasmodium falciparum malaria had a positive association with iNTS disease in three African studies (Table 2) since the year 2000. The estimated OR for iNTS disease in individuals with Plasmodium falciparum malaria compared with those without Plasmodium falciparum malaria ranged from 1.5 to 4.1. In one study, recent history of malaria was positively associated with iNTS disease [192]. Young age was associated with iNTS disease in three studies ( Table 2). The OR for iNTS disease in young individual compared to older individuals ranged from 2.07 to 4.30. In one study of children 1-4 years of age, iNTS disease was reported to be more associated with age when compared to other organisms causing bacteraemia [122]. Anaemia (especially moderate and severe anaemia), was shown by five studies to be associated with iNTS disease in Africa with reported OR ranging from 1.86 to 35.6. Rural settlement compared to urban was found by one study to be associated with iNTS disease [192]. The OR of iNTS disease occurring in a rural settlement compared to an urban settlement was 2.23

Case Fatality Rate (CFR) of iNTS disease in Africa
We were able to extract CFR data for iNTS disease in Africa from twenty four studies (Table 3), describing a total of 548 deaths among 2656 cases. The overall CFR was 20.6%. and ranged from 0% in individuals greater than or equal to 5 years old in Kenya [75] to 72.7% in another Kenyan study involving only HIV-infected patients [81]. The average CFR, derived from 8 studies [3,67,89,93,152,160,172,188] conducted among low risk populations (not HIVinfected, anaemic, malnourished, or having malaria), and with >90 iNTS cases isolated, is 19% (276 fatalities from 1427 cases).

Discussion
Our review seeks to comprehensively document reports of iNTS disease in Africa: the number and location of the reported cases, and where available, the age of the subjects, risk factors, case fatality rates and incidence. As a result, we have included many more studies (199) than a recent review by Ao et al [204] that specifically only included reports of incidence. In their database, in addition to reports from other countries they list 10 reports from 6 African countries. All 10 reports from the Ao et al study are included in the larger set of 14 reports describing incidence from 8 African countries we include in Table 1. Not surprisingly we find a similarly high incidence of iNTS disease in sub-Saharan Africa. Ao et al estimated the overall incidence of iNTS disease in Africa at 227 cases [range 152-341] per 100,000 population with   [204]. We sought to provide a comprehensive review of the published reports of NTS bacteraemia in the peer reviewed literature and found a substantial number of reports with the earliest in 1966, interestingly about the same time as genetic studies suggest the evolution of the highly African-specific highly invasive NTS genotypes [205,206]. There are limits on the comprehensiveness of our study. In particular, many of the early and non-English reports in the initial screening stage were only available to us on line as abstracts, and in some cases, just the title with MESH headings and some of these may have contained details e.g. the use of blood cultures to determine Salmonella bacteraemia in the text but not in the abstract that resulted in these articles not being ordered and reviewed in full. Despite these limitations, this survey highlights important findings.
As shown in Fig 2, no reports were found in some countries in West Africa and in South West Africa. In view of the reported cases from neighbouring countries it seems unlikely that iNTS disease does not occur in these counties, but this highlights that gaps in the published literature almost certainly exist, e.g. due to lack of infrastructure in these countries to undertake the blood culturing required to obtain a definitive diagnosis, or the lack of researchers interested in publishing reports. Some of these gaps in the database may be addressed by examining health service records, but this is beyond the scope of the article.
As shown in Table 1 in the papers we surveyed, the estimated incidence of iNTS disease ranged from 1.4 cases per 100,000 population per year over all age groups in South Africa to a yearly cumulative incidence of 2,520 per 100,000 among <5 years old children in Ghana. However, we caution against trying to over-interpret these incidence data or the range observed-a striking observation is the lack of consistency in the age groups reported, the inclusion criteria that may or may not select for populations with specific risk factors (e.g. HIV, sickle cell anaemia) and in the way in which the population denominator is determined for these largely facility-based studies.
On the basis of the whole database of reported bacteraemia, we confirm the earlier findings that NTS accounts for a large proportion of the bacteria responsible for community acquired blood stream infections and is a substantial cause of morbidity and mortality with the serovars Typhimurium and Enteritidis responsible for 91% of NTS bacteraemia in Africa in the 69% of cases where the serotype was determined.
A major knowledge gap exists in the incidence of iNTS disease in Africa owing to the paucity of reports from population-based surveillance of Salmonella in Africa. This is due to poor surveillance systems for infectious diseases in most parts of Africa. The majority of the studies on iNTS disease in Africa are hospital-based which can only provide an estimate of the minimal incidence of disease in the community, because not all febrile cases report to hospital. Denominators necessary for the detailed estimation of iNTS disease incidence in the reported publications are generally not available. This also makes it difficult to estimate the true population at risk of disease. Population-based incidence studies are best suited to estimate the incidence of an infectious disease in Africa. This can be done by either adjusting the data obtained from hospital-based study with data obtained from simultaneous health-seeking surveys carried out in the same community (for example, as has been carried out by the Typhoid Surveillance in Africa Program (TSAP) and Severe Typhoid in Africa programme [207] or designing a study with a method that will allow for identification of all the cases of the disease within the target community.
Epidemiological data on iNTS disease were available from publications in peer-reviewed journals from 33 of the 54 countries in Africa (Fig 2). Unavailable data from the remaining 21 African countries might be due to a lack of studies in these countries, or a true low burden of iNTS disease. These 21 countries are not involved in TSAP. It would therefore be valuable to conduct a thorough search of grey literature in the affected countries including reviewing existing databases of blood cultures carried out in major hospitals in various locations across the affected countries in the last five to ten years. Such an analysis would facilitate a better understanding of the relative importance of NTS as a cause of community acquired blood stream infection in Africa. New prospective hospital-based blood culture studies in these countries would be even more useful. The average case fatality rate of community acquired severe infections is 20.6% but with a wide range from 0 to 72%. Therefore, the development of an effective vaccine against NTS for Africa would be an important intervention to help reduce the burden of disease and deaths due to NTS. The exact mechanisms of transmission of iNTS disease are currently unclear and there are no rapid diagnostic tests available for its detection. The development of such diagnostics would greatly facilitate the study and management of iNTS disease in Africa.     (11)