+1 Recommend
0 collections
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      Global Burden of Sickle Cell Anaemia in Children under Five, 2010–2050: Modelling Based on Demographics, Excess Mortality, and Interventions

      Read this article at

          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.


          Frédéric Piel and colleagues combine national sickle cell anemia (SCA) frequencies with projected demographic data to estimate the number of SCA births in children under five globally from 2010 to 2050, and then estimate the number of lives that could be be saved following implementation of specific health interventions starting in 2015.

          Please see later in the article for the Editors' Summary



          The global burden of sickle cell anaemia (SCA) is set to rise as a consequence of improved survival in high-prevalence low- and middle-income countries and population migration to higher-income countries. The host of quantitative evidence documenting these changes has not been assembled at the global level. The purpose of this study is to estimate trends in the future number of newborns with SCA and the number of lives that could be saved in under-five children with SCA by the implementation of different levels of health interventions.

          Methods and Findings

          First, we calculated projected numbers of newborns with SCA for each 5-y interval between 2010 and 2050 by combining estimates of national SCA frequencies with projected demographic data. We then accounted for under-five mortality (U5m) projections and tested different levels of excess mortality for children with SCA, reflecting the benefits of implementing specific health interventions for under-five patients in 2015, to assess the number of lives that could be saved with appropriate health care services. The estimated number of newborns with SCA globally will increase from 305,800 (confidence interval [CI]: 238,400–398,800) in 2010 to 404,200 (CI: 242,500–657,600) in 2050. It is likely that Nigeria (2010: 91,000 newborns with SCA [CI: 77,900–106,100]; 2050: 140,800 [CI: 95,500–200,600]) and the Democratic Republic of the Congo (2010: 39,700 [CI: 32,600–48,800]; 2050: 44,700 [CI: 27,100–70,500]) will remain the countries most in need of policies for the prevention and management of SCA. We predict a decrease in the annual number of newborns with SCA in India (2010: 44,400 [CI: 33,700–59,100]; 2050: 33,900 [CI: 15,900–64,700]). The implementation of basic health interventions (e.g., prenatal diagnosis, penicillin prophylaxis, and vaccination) for SCA in 2015, leading to significant reductions in excess mortality among under-five children with SCA, could, by 2050, prolong the lives of 5,302,900 [CI: 3,174,800–6,699,100] newborns with SCA. Similarly, large-scale universal screening could save the lives of up to 9,806,000 (CI: 6,745,800–14,232,700) newborns with SCA globally, 85% (CI: 81%–88%) of whom will be born in sub-Saharan Africa. The study findings are limited by the uncertainty in the estimates and the assumptions around mortality reductions associated with interventions.


          Our quantitative approach confirms that the global burden of SCA is increasing, and highlights the need to develop specific national policies for appropriate public health planning, particularly in low- and middle-income countries. Further empirical collaborative epidemiological studies are vital to assess current and future health care needs, especially in Nigeria, the Democratic Republic of the Congo, and India.

          Please see later in the article for the Editors' Summary

          Editors' Summary


          More than seven million babies are born each year with a structural or functional abnormality. Although some birth defects are caused by environmental factors, many are caused by the inheritance of a defective gene. One common inherited birth defect is sickle cell anemia (SCA). SCA arises when a baby inherits the gene for sickle hemoglobin (HbS), a structural variant of normal adult hemoglobin (HbA, the protein in the disc-shaped red blood cells that carry oxygen round the body), from both its parents. Every cell in the human body contains two full sets of genes, and babies inherit one set of genes from each parent. The parents usually each have one HbS gene and one HbA gene, and are unaffected. However, the red blood cells of their offspring who inherit two copies of HbS develop a sickle (crescent) shape. Sickle cells can block blood vessels in the limbs and organs and have a shorter lifespan than normal red blood cells, which causes anemia. Together, these changes can cause acute pain and organ damage, and can increase the risk of severe infections. SCA can be prevented by prenatal diagnosis and managed by interventions such as the provision of antibiotics and vaccination to prevent infections.

          Why Was This Study Done?

          Without early diagnosis and treatment, children with SCA often die within the first few years of life. Having one copy of the HbS gene provides people with protection from malaria, therefore SCA occurs mainly in low- and middle-income countries in tropical regions, where early diagnosis and treatment is often unavailable. Recent improvements in overall infant and childhood survival in these countries and population migration to higher-income countries mean that the global burden of SCA is likely to increase over the coming decades. To date, no one has tried to quantify this increase, although this information is needed to guide decisions on public health spending. In this modeling study, the researchers assess the size of the expected global burden of SCA between 2010 and 2050 in children under five years old and estimate the number of newborn lives that might be saved by implementation of various health interventions.

          What Did the Researchers Do and Find?

          The researchers used estimates of national SCA frequencies and data on projected birth rates to calculate that the number of newborns with SCA will increase from about 305,800 in 2010 to about 404,200 in 2050. They estimated that Nigeria, the Democratic Republic of Congo (DRC), and India accounted for 57% of newborns with SCA in 2010, and that Nigeria and the DRC will probably still be the countries most in need of policies for the prevention and management of SCA in 2050. The researchers then assessed how many newborns might be saved by the implementation of various health measures in 2015 that affect excess mortality (the difference between the frequency of SCA in newborns and in five-year-olds divided by the frequency of SCA in newborns) among children born with SCA. Implementation of prenatal diagnosis and newborn screening programs, and provision of antibiotics and vaccinations (interventions assumed by the researchers to reduce excess mortality from 90% to 50% in low- and middle-income countries and from 10% to 5% in high-income countries) could prolong the life of more than five million newborns with SCA by 2050. Implementation of universal screening and provision of other specific measures predicted to reduce excess mortality to 5% and 0% in low-to-middle-income countries and high-income countries, respectively, could save nearly ten million lives by 2050.

          What Do These Findings Mean?

          In estimating the global burden of SCA in children under five years old between 2010 and 2050 and the number of newborn lives that could be saved by implementation of health interventions, the researchers made numerous assumptions reflected in the uncertainty associated with the projections. For example, they assumed that implementation of specific interventions would lead to an immediate reduction of excess mortality in newborns with SCA. The study's findings confirm, however, that the global burden of SCA is increasing and indicate that the implementation of specific interventions could extend the lives of millions of newborns with SCA. Although further studies are needed to assess the current and future health care needs of children with SCA, these findings highlight the need to develop and implement national public health planning and funding policies for SCA, particularly in low- and middle-income countries.

          Additional Information

          Please access these websites via the online version of this summary at

          Related collections

          Most cited references 32

          • Record: found
          • Abstract: found
          • Article: not found

          Neonatal, postneonatal, childhood, and under-5 mortality for 187 countries, 1970-2010: a systematic analysis of progress towards Millennium Development Goal 4.

          Previous assessments have highlighted that less than a quarter of countries are on track to achieve Millennium Development Goal 4 (MDG 4), which calls for a two-thirds reduction in mortality in children younger than 5 years between 1990 and 2015. In view of policy initiatives and investments made since 2000, it is important to see if there is acceleration towards the MDG 4 target. We assessed levels and trends in child mortality for 187 countries from 1970 to 2010. We compiled a database of 16 174 measurements of mortality in children younger than 5 years for 187 countries from 1970 to 2009, by use of data from all available sources, including vital registration systems, summary birth histories in censuses and surveys, and complete birth histories. We used Gaussian process regression to generate estimates of the probability of death between birth and age 5 years. This is the first study that uses Gaussian process regression to estimate child mortality, and this technique has better out-of-sample predictive validity than do previous methods and captures uncertainty caused by sampling and non-sampling error across data types. Neonatal, postneonatal, and childhood mortality was estimated from mortality in children younger than 5 years by use of the 1760 measurements from vital registration systems and complete birth histories that contained specific information about neonatal and postneonatal mortality. Worldwide mortality in children younger than 5 years has dropped from 11.9 million deaths in 1990 to 7.7 million deaths in 2010, consisting of 3.1 million neonatal deaths, 2.3 million postneonatal deaths, and 2.3 million childhood deaths (deaths in children aged 1-4 years). 33.0% of deaths in children younger than 5 years occur in south Asia and 49.6% occur in sub-Saharan Africa, with less than 1% of deaths occurring in high-income countries. Across 21 regions of the world, rates of neonatal, postneonatal, and childhood mortality are declining. The global decline from 1990 to 2010 is 2.1% per year for neonatal mortality, 2.3% for postneonatal mortality, and 2.2% for childhood mortality. In 13 regions of the world, including all regions in sub-Saharan Africa, there is evidence of accelerating declines from 2000 to 2010 compared with 1990 to 2000. Within sub-Saharan Africa, rates of decline have increased by more than 1% in Angola, Botswana, Cameroon, Congo, Democratic Republic of the Congo, Kenya, Lesotho, Liberia, Rwanda, Senegal, Sierra Leone, Swaziland, and The Gambia. Robust measurement of mortality in children younger than 5 years shows that accelerating declines are occurring in several low-income countries. These positive developments deserve attention and might need enhanced policy attention and resources. Bill & Melinda Gates Foundation. Copyright 2010 Elsevier Ltd. All rights reserved.
            • Record: found
            • Abstract: found
            • Article: found
            Is Open Access

            Global distribution of the sickle cell gene and geographical confirmation of the malaria hypothesis

            It has been a century since the first description of abnormally elongated red blood cells in an anaemic patient and the link with the clinical symptoms of what is now called sickle cell anaemia (SCA) was published1. Sickle haemoglobin (HbS), a structural variant of normal adult haemoglobin, results from a single amino acid substitution at position 6 of the beta globin molecule (β 6Glu→Val)2. When HbS is inherited from only one parent, the heterozygous child is usually an asymptomatic carrier2. When inherited from both parents, the homozygous child suffers from SCA. HbS is the most common pathological haemoglobin variant worldwide3. Without treatment, which is rarely available in low-income, high-burden countries4, the vast majority of children born with SCA die before the age of 5 years3. Natural selection should therefore have purged this mutation from human populations, but allele frequencies of HbS in excess of 15% have been observed5. In 1949, it was suggested that the Darwinian paradox of high frequencies of genetic blood disorders could result from a selective advantage conferred by such disorders in protecting against Plasmodium falciparum malaria infection in heterozygotes6. This balancing selection, commonly referred to as the 'malaria hypothesis', was originally suggested to explain the geographical correspondence between the distribution of thalassaemia and malaria in the Mediterranean region, and was later confirmed7 in many locations including Sardinia8, Melanesia9 10 and Kenya11. At the same time, a similar relationship between HbS and malaria was independently discovered in Africa12 13. In vitro and in vivo studies have since added support for the protective role of HbS against malaria14 15. Despite significant bibliographic assemblies of information on the distribution of HbS5 16, important limitations exist with previous mapping efforts17 18 19. These include (i) the inclusion of non-random population samples (such as those including patients with malaria or samples from related individuals) that could bias HbS allele frequency estimates; (ii) poor discrimination between indigenous and recently migrated populations that could confound evidence of the relationship between HbS allele frequency and historical malaria endemicity; (iii) the lack of inclusion of HbS allele frequency local geographical heterogeneities; and (iv) limited documentation on the cartographic methodology used to generate maps, making them difficult to reproduce and evaluate objectively. More importantly, the geographical support for the malaria hypothesis has never advanced beyond visual comparison20 21 22 23 24. In this study, we conduct a formal investigation of the geographical evidence in support of the malaria hypothesis at the global scale. In brief, we first updated previous data collections5 16 with online searches of the published literature, which we augmented using unpublished data from the Malaria Genomic Epidemiology Network Consortium (MalariaGEN,, to create a comprehensive geodatabase of HbS allele frequency. These were reviewed using criteria devised to exclude sources of bias, such as those resulting from the inclusion of data from non-representative or non-indigenous populations. We then mapped these data using a Bayesian model-based geostatistical framework26 27 28. This enabled a comparison, for each pixel, between the modelled HbS allele frequency and the endemicity of malaria based on a unique categorical map reflecting its distribution before the era of interventions for malaria control29. Finally, a geostatistical test for geographical association was devised, by computing the areal mean HbS allele frequency associated with each historical malaria endemicity class and calculating the probability that these mean values increased in each successive class. Results HbS allele frequency database and map Searches of the literature identified 41,445 references (see Methods), 90% of which did not include data allowing allele frequency calculations. The application of additional inclusion criteria further restricted the total to 278 informative references (see Supplementary References 64–342, cited in alphabetical order by surname), which have been used as inputs to our model. A total of 699 spatially unique data points were abstracted from these sources and entered into our georeferenced database with 74 additional surveys from MalariaGEN. Of these, 29 (4%) were located in the Americas, 618 (80%) in Africa and Europe (mostly subSaharan Africa) and 126 (16%) in Asia (Fig. 1a and Supplementary Fig. S1). Using our model (see Methods), we produced a continuous 10×10 km resolution global raster grid of HbS allele frequency, with predictions drawn from the median of the posterior predictive distribution for each pixel (Fig. 1b), accompanied by a per-pixel estimate of prediction uncertainty (Fig. 2). Empirical model performance was judged by comparing observed HbS allele frequencies with predicted values for a randomly removed subset of 10% of the data points, which revealed a mean error and a mean absolute error in HbS allele frequency predictions of −0.15 and 6.76%, respectively (see Methods). This global map of HbS allele frequencies should not be interpreted as showing the contemporary geographical distribution of this gene. It is the first global map of the distribution of the HbS gene, based on representative and indigenous population samples (see Methods). Our HbS map (Fig. 1b) showed an HbS allele frequency of >0.5% to be present throughout most of the African continent, the Middle East and India and in localized areas in Mediterranean countries. The maximum predicted value of HbS allele frequency was 18.18% in northern Angola. A large contiguous area with frequencies above 9% was observed stretching from southern Ghana to northern Zambia. The map also indicated similar frequencies in an area extending from southern Senegal to northern Liberia, in localized patches in eastern Côte d'Ivoire, the eastern shores of Lake Victoria, southeast Tanzania and oases on the east coast of Saudi Arabia, as well as in the southern Chhattisgarh and southern Karnataka regions of India. Areas with frequencies above 6% were predicted in Madagascar, central Sudan, the west coast of Saudi Arabia, southeastern Turkey and in the Chalkidiki region of Greece. The many records of absence (Fig. 1a) and the very low HbS allele frequencies predicted by our model (Fig. 1b) also confirmed that HbS was largely absent from the Horn of Africa and from areas south of the Zambezi. Spatial validation of the malaria hypothesis To test the geographical association between HbS and malaria, we used the only available global map of preintervention malaria transmission intensity (endemicity; see Methods)29. On the basis of an assembly of historical malariometric information, this map categorized the world circa 1900 into six classes of successively higher endemicity: malaria free, epidemic, hypoendemic, mesoendemic, hyperendemic and holoendemic (see Fig. 1c for endemicity class definitions)29 30. The relationship between the predicted HbS allele frequencies and the level of malaria endemicity was summarized graphically in violin plots (Fig. 3), which illustrate the density distributions of predicted HbS allele frequencies within each endemic area. HbS was absent from epidemic areas, which were found only in northern America and Eurasia. Globally, predicted HbS allele frequencies were similar in malaria-free, hypoendemic and mesoendemic zones, but were substantially higher in hyperendemic and holoendemic areas (Fig. 3a). In Africa and Europe (Fig. 3b), an increase in HbS allele frequencies from hypoendemic through to holoendemic malaria zones was more pronounced. In Asia (Fig. 3c), no relation between predicted HbS allele frequencies and malaria endemicity was found. HbS was absent in the indigenous populations of the Americas. Although the maps and violin plots provided a valuable insight into the covariation of HbS allele frequency and malaria endemicity, our aim was to formally quantify the significance of any such relationship. Measurement of the difference between the areal mean HbS allele frequency calculated within each endemicity area allowed us to quantify the statistical strength of such differences, taking into account the inherent uncertainty of the predicted HbS allele frequencies (see Methods). Differences in areal means between endemicity regions were calculated for 100 unique realizations of the HbS allele frequency map generated by the Bayesian model (Fig. 4 and Supplementary Fig. S2). When combined, these realizations produced predictive probability distributions for the difference in areal mean HbS allele frequency between each successive endemicity class (see Table 1 and Methods). These geostatistical measures provide the first quantitative evidence for a geographical link between the global distribution of HbS and malaria endemicity. At the global level, we found clear differences between high endemicity classes (Fig. 4a), associated with a high probability of HbS allele frequency increases (>90%) from mesoendemic to hyperendemic and hyperendemic to holoendemic areas, as well as from epidemic to hypoendemic areas (Table 1). In Africa, we observed a gradual increase from epidemic to holoendemic (Fig. 4b). High probabilities of increase were found between the same classes as in the global analysis, but also from hypoendemic to mesoendemic areas (87%). In Asia, differences between classes were much smaller (Fig. 4c) and the probabilities of increase were much lower between most classes, especially in areas of high endemicity. Discussion A strong geographical link between the highest HbS allele frequencies and high malaria endemicity was observed at the global scale (Fig. 4a), but this observation is influenced primarily by the relationship found in Africa (Fig. 4b). The gradual increase in HbS allele frequencies from epidemic areas to holoendemic areas in Africa is consistent with the hypothesis that malaria protection by HbS involves the enhancement of not only innate but also acquired immunity to P. falciparum 31. Interactions with haemoglobin C32 33 might explain the lower HbS allele frequencies in West Africa24. Despite the presence of large malarious areas, HbS is absent in the Americas and in large parts of Asia2 (Fig. 1a). Therefore, no geographical confirmation of the malaria hypothesis could be identified in these regions. Although several haemoglobin variants have been identified in the Americas5, none of the malaria protective polymorphisms have been observed in the indigenous populations of this continent19. The combination of the low likelihood of an independent HbS mutation arising and a relatively low selection pressure (due to the absence of holoendemic areas, the more recent arrival of malaria, as well as the predominance of P. vivax) could contribute to the absence of HbS in that region. In Southeast Asia34, other malaria protective polymorphisms have been identified (haemoglobin E (HbE), the thalassaemias, glucose-6-phosphate dehydrogenase deficiency and Southeast Asian ovalocytosis) and levels of malaria endemicity were relatively high. It is suspected that HbE and Southeast Asian ovalocytosis in particular may have had epistatic interactions35 36, altering the selection pressure for the HbS gene in that region37. The complex social structure and the predominance of P. vivax 38 are also considered as likely to contribute to the unresolved geographical relationship in India. Ongoing work to create an open-access database for several malaria protective polymorphisms will allow more comprehensive distribution mapping and improve our understanding of their geographical interaction. Substantial variations in HbS allele frequencies over short distances (up to 10% over <50 km) have been described in literature5, for example, in relation to altitude, rainfall or Anopheles survival39, which underlie variations in selection40. Such spatial heterogeneity was observed in the geodatabase. The combination of a detailed georeferencing process, the use of a geostatistical model able to incorporate the multiple scales of variation within the data and a semicontinuous gradient of HbS allele frequencies allowed us to describe the global distribution and the high geographical variability of this gene more rigorously than achieved in previous maps17 18 19. The uncertainty measure (see Fig. 2) provides an important estimate of the limitations associated with a retrospective data set, and can highlight areas prone to small population samples and/or areas lacking observations (for example, New Zealand). Among the factors that might contribute to the heterogeneity observed in the HbS allele frequency in hyperendemic areas in Africa (Fig. 3b), we identified (i) a component of geographical sampling error from an 'opportunistic sample' of surveys that we were able to source from literature; (ii) the kinetics of the spread of the HbS gene, which leads to an exponential increase in areas in which a selective pressure appears, but to a much slower decrease in areas in which the selective pressure disappears41; (iii) long-term (sociological or physical) isolation of local populations, which could result in pockets of lower HbS allele frequencies observed on the map (Fig. 1b). One hundred years after the first description of SCA, we used a comprehensive search combined with a rigorous selection of survey data and modern mapping methods to create a new, evidence-based map of the worldwide distribution of HbS allele frequency and to quantify the uncertainty in these mapped predictions. Using a novel geostatistical approach that accounts for this uncertainty, we have compared this new map with a historical map of the global endemicity of malaria. We provide the first geographical and quantitative confirmation of the malaria hypothesis at the global scale. Methods Creating a global database of sickle cell allele frequencies A schematic overview of the methods used is provided as Figure 5. To identify publications with HbS allele frequency data, a comprehensive electronic data search was undertaken using PubMed (, ISI Web of Knowledge ( and Scopus (, using the following keyword string: 'sickle cell' or 'haemoglobin S' or 'hemoglobin S' or 'Hb S'. Initial searches were conducted on 12 December 2007 and updated on 20 October 2009. A total of 18,336 (in Text terms), 28,908 (in Title/Keywords/Abstract) and 22,732 (in Article Title/Abstract/Keywords) references were found in the three respective databases and exported using bibliographic management software. The 2,220 references from Livingstone's extensive but out-of-date database on frequencies of haemoglobin variants5 were then added. Duplicates were removed manually. Titles and abstracts, when available, were then reviewed to identify references that met the following selection criteria: first, that the population sample was representative of an indigenous population. When multiple surveys included similar subsets of population samples, only the larger one was included, provided that all the other inclusion criteria were fulfilled. When multiple surveys were totally independent, each survey was included in the model. Few studies corresponded to a purely random or universal sample of the population studied; therefore, all unselected samples were included. Studies of patients, with sickle cell or any other condition, were excluded. We considered population surveyed as indigenous, if no information was available from the author to suspect that the population did not evolve locally in relation to the historical prevalence of malaria. Non-native populations surveyed in the Americas or Western Europe, for example, were therefore excluded. Surveys explicitly surveying a specific ethnic group, not representative of the overall population at the sampling site, were excluded. Although ethnic group information was recorded when available, it was not used in the model because of (i) inconsistency of information provided by the sources and ethnic group definitions used and (ii) contradicting local results in the relationship between ethnicity and HbS allele frequency. Second, details were needed on the number of individuals sampled and on the AA and AS genotypes identified. Sources reporting an allele frequency but no sample size were thus excluded. Because of (i) the complexity of the multiple compound status when HbS is inherited with another structural variant, haemoglobin C or HbE, or with a thalassaemia, α- or β-, (ii) the small number of individuals involved (apart from in the Mediterranean countries) and (iii) the inconsistencies in the identification of such cases, these individuals were not included in the calculations of the HbS allele frequency. Third, the survey description needed to be spatially explicit so that it could be georeferenced (see below). Using these strict criteria for inclusion, we identified 278 references with data allowing us to calculate an allele frequency for HbS (see Supplementary References 64–342). Data on absences of HbS in populations, such as native Americans, were also included in this study, as they usually constituted isolated data points that are very informative for a global predictive model. Finally, genotype data collected by the Malaria Genomic Epidemiology Network Consortium (MalariaGEN, were added to the database as they represent a significant source of standardized data from malaria-endemic countries. Georeferencing We used the georeferencing procedure developed by the Malaria Atlas Project (MAP,, which is described in Guerra et al.42 Geographic coordinates could be found for 459 population samples, located as points (<10 km2). The centroid of polygons was used for the 314 population samples that could be georeferenced to district level (admin2 unit) or to a smaller area clearly defined by the author (for example, detailed map of the study area). Studies that could only be located to province (admin1 unit) or country (admin0 unit) level were excluded. Creating a continuous map of sickle cell allele frequency The number of individuals with AA and AS genotypes was used to calculate allele frequencies. Individuals described as sicklers were all considered as heterozygotes (AS). All SS individuals were assumed to die shortly after birth, meaning we discarded the few records of SS individuals in the database. Preliminary analysis (not shown) indicated that the resulting likelihood functions at points with SS individuals were very similar to those obtained using standard Hardy–Weinberg assumptions. Even today, medical services for improving the survival of sickle cell patients (SS) are rarely available outside economically developed countries, where the burden of sickle cell is greatest, and would have been more rudimentary before the 1990s, when two-thirds of the surveys were conducted. It seems reasonable therefore to assume that the few surviving HbS homozygous individuals were unlikely to substantively affect HbS allele frequency estimates. When only an allele frequency and the sample size were given, the number of AA and AS individuals was calculated by assuming that the genotypes of newborns were in Hardy–Weinberg proportions but that all SS individuals had died by the time of the surveys. The sample size was recalculated as the sum of AA and AS individuals. Information on age could not be taken into account as it was provided in only 45% of the sources. Among these, samples were taken from cord blood/neonates (n=23,152), children (n=26,205), adults (n=219,966) and mixed groups (n=78,111). The inputs to the geostatistical model were the coordinates of the population studied (lat/long in decimal degrees, WGS84) and the number of AS (positive) and AA (negative) individuals. A Bayesian geostatistical model involving a two-part nested covariance function was fitted to these data and 500,000 Markov chain Monte-Carlo iterations43 were used to predict HbS allele frequencies at unsampled locations and generate continuous maps. Because of the high heterogeneity of allele frequencies in areas in which HbS is present, the small set of HbS absences, for example, in the Americas, was not sufficient to rule out the possibility that HbS allele frequency could be relatively high in some places. For that reason, the posterior predictive distribution of allele frequencies in the Americas tended to have a long right-hand tail, and point estimates of allele frequency tended to be surprisingly high. A thinned 10% sample of the data was used to map various summary statistics of the posterior predictive distribution of HbS allele frequency at unsampled locations. To validate the predictions of the model, the analysis was repeated with 90% of the data set, and predictions at the locations of the held-out data points were evaluated. See Supplementary Methods for details on the statistical analysis. Comparing with a precontrol map of malaria endemicity In the late 1960s, a team of Russian researchers conducted a synthesis of historical records, documents and maps of several malariometric indices used to record malaria endemicity29. Combined with expert opinion and data on temperature and rainfall, this review allowed them to create a unique global map of the precontrol distribution of malaria, at the peak of its hypothesized distribution44. We chose to use this malaria map for reasons detailed in Supplementary Methods. Similar to traditional box plots, violin plots allow the comparison of a semicontinuous variable (HbS allele frequency) with a categorical variable (malaria endemicity class). In addition, they show the density distribution of the observations or predictions. The analysis of the violin plots of the allele frequencies within each endemicity class is supported by a visual interpretation of the plots. To quantify the differences between malaria endemicity classes, we calculated the probability of finding a higher allele frequency in one class than in the class just below on the basis of their geographical pattern. The posterior predictive distribution of the areal mean of the HbS allele frequency over each endemicity class was plotted by region (Fig. 4). The posterior probability of an increase in the areal mean HbS allele frequency for each pair of consecutive malaria endemicity classes, along with the Monte-Carlo standard errors associated with those estimates, was then calculated. Probabilities of zero and one indicate that the HbS allele frequency in an endemicity class is certainly lower or higher, respectively, than in the adjacent class. A probability of 0.5 corresponds to an equal chance of an increase or decrease. Further details are provided in Supplementary Methods. These comparisons have been made globally and regionally for Europe and Africa, and for Asia. The separation into these two regions was based on the distinct haplotypes occurring east and west of Saudi Arabia (see Fig. 1a)19 37. Author contributions F.B.P. and S.I.H. helped to assemble the data, developed the conceptual approach and wrote the first draft of the manuscript. R.E.H. and O.A.N. assembled and abstracted the data. A.P.P. and P.W.G. conceived and helped to implement the modelling and all computational tasks. All authors contributed to the study design and data interpretation and to the revision of the final manuscript. Additional information How to cite this article: Piel, F.B. et al. Global distribution of the sickle cell gene and geographical confirmation of the malaria hypothesis. Nat. Commun. 1:104 doi: 10.1038/ncomms1104 (2010). Supplementary Material Supplementary Figures, Supplementary Methods, Supplementary References Supplementary Figures S1–S2, Supplementary Methods and Supplementary References
              • Record: found
              • Abstract: found
              • Article: not found

              Bacteraemia in Kenyan children with sickle-cell anaemia: a retrospective cohort and case–control study

              Summary Background In sub-Saharan Africa, more than 90% of children with sickle-cell anaemia die before the diagnosis can be made. The causes of death are poorly documented, but bacterial sepsis is probably important. We examined the risk of invasive bacterial diseases in children with sickle-cell anaemia. Methods This study was undertaken in a rural area on the coast of Kenya, with a case–control approach. We undertook blood cultures on all children younger than 14 years who were admitted from within a defined study area to Kilifi District Hospital between Aug 1, 1998, and March 31, 2008; those with bacteraemia were defined as cases. We used two sets of controls: children recruited by random sampling in the same area into several studies undertaken between Sept 1, 1998, and Nov 30, 2005; and those born consecutively within the area between May 1, 2006, and April 30, 2008. Cases and controls were tested for sickle-cell anaemia retrospectively. Findings We detected 2157 episodes of bacteraemia in 38 441 admissions (6%). 1749 of these children with bacteraemia (81%) were typed for sickle-cell anaemia, of whom 108 (6%) were positive as were 89 of 13 492 controls (1%). The organisms most commonly isolated from children with sickle-cell anaemia were Streptococcus pneumoniae (44/108 isolates; 41%), non-typhi Salmonella species (19/108; 18%), Haemophilus influenzae type b (13/108; 12%), Acinetobacter species (seven of 108; 7%), and Escherichia coli (seven of 108; 7%). The age-adjusted odds ratio for bacteraemia in children with sickle-cell anaemia was 26·3 (95% CI 14·5–47·6), with the strongest associations for S pneumoniae (33·0, 17·4–62·8), non-typhi Salmonella species (35·5, 16·4–76·8), and H influenzae type b (28·1, 12·0–65·9). Interpretation The organisms causing bacteraemia in African children with sickle-cell anaemia are the same as those in developed countries. Introduction of conjugate vaccines against S pneumoniae and H influenzae into the childhood immunisation schedules of African countries could substantially affect survival of children with sickle-cell anaemia. Funding Wellcome Trust, UK.

                Author and article information

                Role: Academic Editor
                PLoS Med
                PLoS Med
                PLoS Medicine
                Public Library of Science (San Francisco, USA )
                July 2013
                July 2013
                16 July 2013
                : 10
                : 7
                [1 ]Evolutionary Ecology of Infectious Disease, Department of Zoology, University of Oxford, Oxford, United Kingdom
                [2 ]Spatial Ecology and Epidemiology Group, Department of Zoology, University of Oxford, Oxford, United Kingdom
                [3 ]Global Network for Sickle Cell Disease, Toronto, Ontario, Canada
                [4 ]Weatherall Institute of Molecular Medicine, University of Oxford, Oxford, United Kingdom
                [5 ]Kenya Medical Research Institute/Wellcome Trust Programme, Centre for Geographic Medicine Research-Coast, Kilifi District Hospital, Kilifi, Kenya
                [6 ]Department of Medicine, Imperial College, St Mary's Hospital, London, United Kingdom
                Institute for Global Health, United Kingdom
                Author notes

                The authors have declared that no competing interests exist.

                Conceived and designed the experiments: FBP. Performed the experiments: FBP. Analyzed the data: FBP TNW. Wrote the first draft of the manuscript: FBP. Contributed to the writing of the manuscript: FBP SIH SG DJW TNW. ICMJE criteria for authorship read and met: FBP SIH SG DJW TNW. Agree with manuscript results and conclusions: FBP SIH SG DJW TNW.


                This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

                Page count
                Pages: 1
                This study was supported by funding from the Wellcome Trust (Biomedical Resources Grant #085406, PI: SIH) and the European Research Council (Advanced Grant - DIVERSITY, PI: SG). SIH is funded by a Senior Research Fellowship from the Wellcome Trust (#095066). TNW is funded by a Senior Clinical Fellowship from the Wellcome Trust (#091758). The funding agencies had no role in the design and conduct of the study; in the collection, management, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript. This paper is submitted with permission of the Director of KEMRI.
                Research Article
                Human Genetics
                Genetic Counseling
                Population Genetics
                Earth Sciences
                Human Geography
                Spatial Analysis
                Disease Mapping
                Genetic Epidemiology
                Spatial Epidemiology
                Global Health
                Sickle Cell Disease
                Public Health
                Child Health
                Health Screening
                Preventive Medicine



                Comment on this article