Global epidemiology of sickle haemoglobin in neonates: a contemporary geostatistical model-based map and population estimates

Summary

Background

Reliable estimates of populations affected by diseases are necessary to guide efficient allocation of public health resources. Sickle haemoglobin (HbS) is the most common and clinically significant haemoglobin structural variant, but no contemporary estimates exist of the global populations affected. Moreover, the precision of available national estimates of heterozygous (AS) and homozygous (SS) neonates is unknown. We aimed to provide evidence-based estimates at various scales, with uncertainty measures.

Methods

Using a database of sickle haemoglobin surveys, we created a contemporary global map of HbS allele frequency distribution within a Bayesian geostatistical model. The pairing of this map with demographic data enabled calculation of global, regional, and national estimates of the annual number of AS and SS neonates. Subnational estimates were also calculated in data-rich areas.

Findings

Our map shows subnational spatial heterogeneities and high allele frequencies across most of sub-Saharan Africa, the Middle East, and India, as well as gene flow following migrations to western Europe and the eastern coast of the Americas. Accounting for local heterogeneities and demographic factors, we estimated that the global number of neonates affected by HbS in 2010 included 5 476 000 (IQR 5 291 000–5 679 000) AS neonates and 312 000 (294 000–330 000) SS neonates. These global estimates are higher than previous conservative estimates. Important differences predicted at the national level are discussed.

Interpretation

HbS will have an increasing effect on public health systems. Our estimates can help countries and the international community gauge the need for appropriate diagnoses and genetic counselling to reduce the number of neonates affected. Similar mapping and modelling methods could be used for other inherited disorders.

Funding

The Wellcome Trust.

Introduction

Sickle haemoglobin (HbS) is the most common pathological haemoglobin mutation worldwide.¹ This mutation is a structural variant of normal adult haemoglobin (HbA), which is inherited as a Mendelian trait. Carriers, or heterozygotes (AS individuals), inherit an HbS allele from one parent and an HbA allele from the other. These individuals are usually asymptomatic. Homozygotes (SS individuals) who have inherited HbS alleles from both parents suffer from sickle cell anaemia, which often leads to acute and chronic complications.² Without treatment, which is rarely available in low-income high-burden countries, it is assumed that most children born with the disease die in their first years of life.³

A map showing the global distribution of the HbS allele among indigenous populations has been published⁴ and used for geographical confirmation of the malaria hypothesis, which suggests that the gene responsible for sickle haemoglobin disorders could reach high frequencies because of the resistance conferred against malaria by the heterozygous carrier state.⁵ This map does not, however, reflect the contemporary distribution of the populations with the HbS allele. Following human diasporas,⁶ the HbS allele has spread from the highly malarious regions (where it was selected for the resistance that it confers to malaria) to most other regions worldwide.⁷ In low-income and middle-income countries, the epidemiological transition resulting from economic development and improvements in hygiene, nutrition, and public health (including malaria control) has contributed to significantly reduced infant and child mortality.⁸ As a consequence, patients with HbS who would have previously died undiagnosed in their first years of life can now survive into adulthood.⁹ Despite substantial decreases in malaria endemicity worldwide¹⁰ and hence reduced selection pressure for the HbS gene, population dynamics dictate that it would take many generations for HbS to reach allele frequencies close to zero (for example 100 generations for an initial frequency of 0·06 and fitness values of 1·0 and 0 for AS and SS individuals, respectively).¹¹ Additionally, the population growth rate is still relatively high in many countries with high frequencies of HbS so that, without prenatal diagnosis and avoidance strategies,¹² existing allele frequencies will lead to a progressive increase in the absolute number of SS neonates for many generations. In high-income countries, screening programmes are now commonly needed to diagnose HbS in mixed populations to provide counselling and to prevent serious clinical complications.¹³ These factors strongly suggest that the number of AS and SS cases and associated care costs will substantially increase over the coming decades.^12,14–16

In this context, reliable estimates of the populations affected by haemoglobin disorders are desirable to guide public-health priority settings. We aimed to develop a new cartographic method to provide global, regional, national, and subnational (in data-rich areas) estimates of AS and SS neonates accompanied by appropriate uncertainty measures. Our study generated the following products: (1) a contemporary georeferenced database of HbS allele frequency surveys, which will be made freely accessible online as part of the Malaria Atlas Project's Repository of Open Access Data (ROAD-MAP); (2) a Bayesian model-based geostatistical framework¹⁷ that uses spatial data to predict allele frequencies in unsampled areas and produces a continuous global map of the contemporary distribution of HbS; (3) global, regional, national, and subnational annual estimates of HbS neonate heterozygotes and homozygotes with areal uncertainty predictions,¹⁸ based on Hardy-Weinberg equilibrium^19,20 and on the pairing of the allele frequency predictions with demographic data.

Methods

Database and contemporary map of HbS allele frequency

Online bibliographic searches were done to create a database of spatially located surveys of HbS allele frequency, supplemented by various unpublished datasets (figure 1). The temporal distribution and sample size of surveys are shown in the appendix pp 4–10. We excluded community surveys for which any selection bias (eg, ethnicity or health status) in the sampling methodology was stated or suspected. Remaining surveys were judged to be representative of the local populations and retained only if the sample size and the number of AS individuals (and SS individuals when given) were reported and they could be georeferenced to the district level (second administrative level) or below.²¹ The input data for the model were geographic coordinates (latitude and longitude in decimal degrees) and the number of HbS and non-HbS alleles in the population samples. A Bayesian geostatistical model, described in detail previously,⁴ was fitted to the data and used to predict HbS allele frequencies at unsampled locations and generate continuous maps at 5×5 km resolution (appendix pp 10–15). Differences between these maps and the ones previously published⁴ are further highlighted in the appendix pp 17. Validation metrics were calculated by comparing the observed allele frequency for a 10% random hold-out sample (a subset of the data removed from the validation analysis, so that the observed values can be compared with the predicted values) of the dataset with the prediction output created from the thinned 90% subset of the data (appendix pp 15–17).

Figure 1.

Schematic overview of the methods and Bayesian model-based geostatistical analysis

HbS=sickle haemoglobin. Positive=number of HbS alleles in the population sample. Negative=number of non-HbS alleles in the population sample. Mbg-infer=model-based geostatistical inference process based on a Markov chain Monte Carlo (MCMC) algorithm. GRUMP=Global Rural Urban Mapping Project. Mbg-map=model-based geostatistical mapping process. Mbg-areal-predict=model-based geostatistical areal prediction process. AS=HbS heterozygotes. SS=HbS homozygotes.

Defining estimates of neonates affected in 2010

We estimated the number of neonates affected in 2010 by combining our HbS allele frequency predictions with high-resolution population distribution data (figure 1) from the Global Rural Urban Mapping Project (GRUMP).²² For each country, the crude birth rate for 2010–15 was downloaded from the 2010 online revision population database of the UN world population prospects²³ and combined with the population surfaces to estimate the total number of neonates expected in 2010 globally within each 5×5 km pixel (appendix pp 19–20).

Aggregating per-pixel HbS allele frequency predictions to subnational (for national first-level subdivisions [ie, province or district, containing at least two datapoints]), national, regional, or global scales, while maintaining the correct statistical-dependence structure and therefore allowing these estimates to be calculated with appropriate credible intervals (CIs)¹⁸ required a bespoke version of the Bayesian framework developed previously.⁴ Here we used simulations from the joint posterior of the model parameters and HbS allele frequency at the observation locations to generate a posterior predictive distribution (PPD) for the HbS frequency surface within each area of interest (figure 2). Assuming Hardy-Weinberg equilibrium^19,20 with two alleles (HbA and HbS), if x is the HbS allele frequency at a given pixel, x² and 2x (1–x) represent SS and AS frequencies, respectively. By integrating samples of these quantities from their PPDs over the area of interest, weighted by the population density and crude birth rate, we produced a PPD for the number of neonates affected, from which various summary estimates (eg, posterior mean or median) and CIs (eg, 95%) could be generated (appendix pp 20–22).

Figure 2.

Datapoint distribution and maps of the mean and uncertainty in the predicted HbS allele frequency

(A) Distribution of the datapoints. Red circles and blue triangles indicate surveys showing presence and absence of HbS, respectively. (B) Mean of the posterior predictive distribution. (C) Bayesian model-based geostatistics prediction uncertainty (posterior standard deviations) of the HbS allele frequency.

WHO regional country groupings used previously⁷ might not be the most useful for summarising HbS epidemiology, because of the variable clinical severity of the multiple HbS haplotypes²⁴ and its link with malaria endemicity.²⁵ We therefore present our results both within WHO regions and within modified regional groupings, called HbS regions. The rationale for using modified regions is described in more detail in the appendix pp 22–24.

Results

A total of 1211 spatially unique datapoints (figure 2) from 435 references (web dataset) matched all our inclusion criteria. Of these, 51% (619) were located in the African continent, 23% (273) in the Asian continent, 17% (203) in the Americas, 9% (112) in Europe, and <1% (four) in Oceania. Absence of HbS in the populations tested was recorded for 292 datapoints, of which 117 were in Asia, 86 in Africa, 69 in the Americas, 17 in Europe, and three in Oceania. 9 032 377 individuals were included in the selected surveys with overall sample sizes ranging from one to 3 212 374 individuals (appendix p 5). The mean overall sample size was 7458 (SD 105 542) individuals.

To our knowledge, the map presented here, illustrated by the mean of the PPD (figure 2), is the first cartographic representation of the contemporary distribution of sickle haemoglobin globally. Although frequencies above 20% have been reported in small samples (<100) of specific ethnic groups, frequencies above 18% are unlikely according to the model at the population level. The maximum predicted median allele frequency was only identified in one small patch in northern Angola. Predictions above 15% were commonly identified in that area, as well as around the Gabon Estuary (Libreville) and in localised areas in southwestern Nigeria (Ibadan) and east of Lake Victoria. Predictions between 10·0% and 12·5% were found around the areas described above, in two patches in the south of the DR Congo and in southern Karnataka, India. Frequencies between 7·5% and 12·5% were predicted in large areas extending from southern Senegal to northern Liberia, and from southern Ghana to northern Zambia. Smaller areas of similar frequencies were predicted in Belize; in Khalkidhiki, Greece; in north-eastern Côte d’Ivoire, northwestern Egypt, south of Lake Chad, in central Sudan, around Lake Victoria and the coast of Kenya, Tanzania, and Mozambique, and on the east coast of Madagascar; in southern Turkey, southwestern Syria and eastern Saudi Arabia; and in Karnataka, Chhattisgarh, and Orissa in India. The maximum predicted frequency in South America was about 5%. Predicted median frequencies in western Europe and localised patches in North America (California, Mississippi, or Alabama) did not exceed 1%. By contrast with the high HbS allele frequencies predicted across most of sub-Saharan Africa, large areas where HbS is generally absent were identified in the south and Horn of Africa, as well as in Morocco and Algeria.

The highest uncertainties in predicted allele frequencies were seen in the western area of the DR Congo and in northwestern Angola, where predictions ranged between 3% and 14%, corresponding to an IQR of 11% (figure 2). Relatively high uncertainties (IQR ≥7·5%) were seen across most of sub-Saharan Africa and in parts of Egypt and Libya; in eastern Cuba, Nicaragua and Honduras; in eastern Saudi Arabia; and in Karnataka, India. The maximum IQRs in North America and South America were 5% and 7%, respectively. Uncertainty up to 7% was also seen in parts of Greece (1–8%), Syria (0–7%, 1–8%, and 2–9%, in different parts of the country), and Lebanon (2–9%).

Comparison of the PPD predicted from the thinned dataset with the known values at the locations from the hold-out subset revealed a minimal overall bias, with a mean error in allele frequency prediction of 0·2%. The mean absolute error, which measures the average magnitude of allele frequency prediction errors, was 3·0%. The predicted allele frequencies are therefore only slightly overestimated (ie, mean errors were mostly small and positive), but predicted HbS allele frequencies could be substantially different from the observed values (ie, absolute errors up to 12·5% locally), especially in areas with a high short-range heterogeneity in the data and where the distribution of the dataset was patchy (appendix pp 15–17).

Pairing demographic data with our predicted allele frequencies, we estimated that, in 2010, 312 302 (IQR 294 307–329 729) homozygous SS babies and 5 476 407 (IQR 5 290 779–5 679 288) heterozygous AS babies were born globally (table 1); the total number of neonates affected was 5 788 709 (IQR 5 585 086–6 009 017). At the regional scale, 64·4% of the AS neonates were from sub-Saharan Africa, 22·7% in the Arab-India region, 7·4% in the Americas, 5·4% in Eurasia, and 0·1% in southeast Asia. Among the SS neonates, we estimated that 75·5% were born in sub-Saharan Africa (table 2).

Table 1.

Summary of global annual predicted estimates of neonates with the HbS allele

	Mean (SE)*	Median (SE)*	IQR (SE)*†	Modell and Darlison†
HbAS	5 496 253 (27 780)	5 476 407 (30 783)	5 290 779 (25 205) to 5 679 288 (21 244)	..
HbSS	313 945 (2008)	312 302 (1779)	294 307 (1180) to 329 729 (2401)	222 785‡

Population 6 912 202 000 (crude birth rate 0·01899²³).

^*Monte Carlo standard error.

^†Neonate estimates from Modell and Darlison.⁷

^‡Estimate falling outside our 90% credible interval.

Table 2.

Summary of regional annual predicted estimates of HbAS and HbSS neonates

	Population*	Crude birth rate	HbAS neonates/year			HbSS neonates/year
			Mean	Median (IQR)	%	Mean	Median (IQR)	%	Modell and Darlison†
WHO regions
AFRO	888 817	0·0357	3 607 022	3 610 851 (3 498 595–3 704 303)	64·2	239 547	238 083 (224 003–253 047)	75·4	184 812‡
AMRO	939 833	0·0162	398 279	391 257 (358 199–435 894)	7·6	13 708	13 104 (11 126–15 606)	4·6	4432‡
EMRO	560 803	0·0249	275 365	256 643 (199 839–327 983)	5·7	10 007	8239 (6012–11 951)	3·6	7389
EURO	893 002	0·0123	127 494	121 601 (99 414–147 505)	2·6	3653	3271 (2408–4366)	1·3	376‡
SEARO	1 789 082	0·0200	1 040 033	1 020 489 (900 452–1 154 480)	20·0	44 132	42 597 (35 022–50 750)	15·1	25 768‡
WPRO	1 840 667	0·0128	2292	1150 (477–2374)	0·0	4	9 (2–33)	0·0	9
HbS regions
Americas	939 724	0·0162	389 892	386 430 (349 253–425 791)	7·4	13 309	12 802 (10 869–15 210)	4·6	..
Arab-India	1 771 305	0·0219	1 168 805	1 147 477 (1 010 443–1 299 147)	22·7	48 951	46 826 (39 147–56 000)	16·9	..
Eurasia	1 098 104	0·0139	271 474	256 163 (216 499–310 758)	5·4	8784	7493 (5919–10 090)	3·0	..
Southeast Asia	2 215 004	0·0133	4854	2535 (1324–5171)	0·1	80	21 (7–63)	0·0	..
Sub-Saharan Africa	888 065	0·0365	3 579 982	3 580 207 (3 473 117–3 684 718)	64·4	237 253	235 681 (220 993–250 568)	75·5	..

Crude birth rate data obtained from the UN world population prospects.²³ Estimates are shown by WHO regions and HbS regions. See appendix for further details. AFRO=Regional Office for Africa. AMRO=Regional Office for the Americas. EMRO=Regional Office for the Eastern Mediterranean Countries. EURO=Regional Office for Europe. SEARO=Regional Office for South-East Asia. WPRO=Regional Office for the Western Pacific.

^*Population data in thousands.

^†SS neonate estimates from Modell and Darlison.⁷

^‡Estimate falling outside our 90% credible interval.

We calculated that 50% of the total AS and SS neonates were born in only three countries: Nigeria (1 223 330 [IQR 1 146 733–1 295 851]), India (1 038 579 [918 557–1 186 209]), and DR Congo (527 963 [488 571–568 883]). As the heterozygotes represented 94% of this total, the ranking is the same for the AS neonates, with an estimated 1 136 909 (IQR 1 070 320–1 200 907) in Nigeria, 996 563 (883 210–1 135 290) in India, and 489 745 (455 733–524 014) in DR Congo. For SS neonates, the estimates for these countries were 85 186 (IQR 76 413–94 945) in Nigeria, 42 016 (35 347–50 919) in India, corresponding to 88% of the homozygous cases in Asia (Arab-India and southeast Asia), and 38 217 (32 838–44 870) in DR Congo (figures 3–5, appendix pp 29–38).

Figure 3.

Comparison plots of predicted national SS neonate estimates with Modell and Darlison's estimates for the AFRO and AMRO regions

Red dots show our estimates (termed MAP) with IQRs shown as red lines. Blue dots represent Modell and Darlison's estimates.⁷ (A) AFRO. (B) AMRO. AFRO=Regional Office for Africa. AMRO=Regional Office for the Americas.

Figure 4.

Comparison plots of predicted national SS neonate estimates with Modell and Darlison's estimates for the EMRO and EURO regions

Red dots show our estimates (termed MAP) with IQRs shown as red lines. Blue dots represent Modell and Darlison's estimates.⁷ (A) EMRO. (B) EURO. EMRO=Regional Office for the Eastern Mediterranean Countries. EURO=Regional Office for Europe.

Figure 5.

Comparison plots of predicted national SS neonate estimates with Modell and Darlison's estimates for the SEARO and WPRO regions

Red dots show our estimates (termed MAP) with IQRs shown as red lines. Blue dots represent Modell and Darlison's estimates.⁷ (A) SEARO. (B) WPRO. SEARO=Regional Office for South-East Asia. WPRO=Regional Office for the Western Pacific.

Areal estimates calculated at the subnational level for data-rich areas showed principal administrative units (eg, provinces or districts) in which the predicted number of neonates affected were the highest (appendix pp 39–44). In DR Congo, we estimated that about a third of SS births occurred in three provinces each year: Equateur (3730 [IQR 2805–5019]), Katanga (5514 [4175–7350]), and Province Orientale (3711 [2945–4687]). In Nigeria, about 20% of the total number of SS neonates was predicted in the states of Lagos (8457 [7047–10 143]), Oyo (5315 [4520–6272]), and Kaduna (3722 [2956–4624]). In India, 25% of the SS neonates were predicted in the Karnataka state (9394 [6936–12 890]). Together, Tamil Nadu (5221 [3271–8169]) and Maharashtra (4678 [3511–6644]) represented about another 25% of the national SS prediction. Complete subnational estimates for data-rich areas are shown in the appendix pp 39–44.

Discussion

Sickle cell anaemia has recently been recognised as a global public health problem by the UN and WHO.²⁶ The burden of this disorder and other haemoglobin disorders is expected to increase in the coming decades, because of reductions in infant mortality in many low-income countries and increasing migration from high to low HbS-frequency areas worldwide.¹² A detailed assessment of our current epidemiological knowledge and of the magnitude of the problem is needed to define adequate policies for diagnostics, education, and treatment to prevent or manage complications associated with sickle cell anaemia. We believe that the first contemporary global map of the distribution of sickle haemoglobin (figure 2) and derived global, regional, national, and subnational annual estimates with given precision of AS and SS neonates presented in our report represent substantial contributions towards this goal (tables 1, 2; appendix pp 29–44; panel). The best way to reliably assess the situation would generally be through universal national neonate screening.²⁹ However, such programmes are rare, especially in areas with high HbS frequencies (appendix pp 25–26). Using existing survey data, combined with bespoke mapping and modelling methods, our aim was not only to update previous estimates⁷ but also to take into account the striking geographical variation both in the availability of survey data, and in the frequency of HbS between and within populations, and to assess the reliability of our estimates.

Panel. Research in context.

Systematic review

We searched PubMed, ISI Web of Knowledge, and Scopus, for articles including epidemiological data on HbS allele frequency. We used strict inclusion criteria regarding the sampling and diagnostic methods, as well as the precision of the study location, but no restriction was applied for the publication year or the sample size. We also contacted study investigators and experts in the field either to clarify some information or to complement our database with unpublished data. All selected surveys were included in our global database. Our Bayesian geostatistical model accounted for uncertainty associated with the input data and produced precision measures. Further details are given in the appendix pp 3–18.

Interpretation

Few studies have attempted to quantify the burden of HbS. In 2001, Weatherall and colleagues¹ estimated that about 216 000 babies were born with sickle cell anaemia in Africa, and between 60 000 and 100 000 outside of Africa. In 2007, using a partially updated version of Livingstone's database,²⁷ Modell and Darlison⁷ calculated global, regional, and national annual estimates of babies born with sickle cell anaemia, but only some results have been published. Moreover, the method was based on crude national averages of datasets assembled with less strict inclusion criteria; the precision of these estimates is unknown. Using data from the 2003 United Nations Demographic Yearbook, Modell and Darlison estimated that 222 785 babies were born with sickle cell anaemia globally.^7,28 Their regional and national estimates are shown in table 2 and figures 3–5, respectively. Our study is, therefore, the first rigorous evidence-based analysis focusing on sickle haemoglobin and using a contemporary dataset combined with demographic data and modern geostatistical modelling techniques that account for spatial heterogeneities and provide precision measures.

Our map shows that high frequencies of HbS (>10%) are almost exclusively encountered in sub-Saharan Africa (figure 2). The areas of highest frequency in Africa seem to reflect the origins of the different African haplotypes (Senegal, Benin, and Central African Republic or Bantu).³⁰ Frequencies of HbS seem highly heterogeneous across the Indian subcontinent. HbS is still rare in southeast Asian populations. In the Americas, the distribution of HbS, which is common in populations along the eastern coast of the continents, mirrors the distribution of African Americans. Uncertainties were highest in areas for which no data were available, such as central DR Congo (figure 2). Data in those areas would be particularly useful to improve our knowledge of the distribution of this gene. Our validation process showed that our model performed well in view of the amount of data available.

Micromapping studies have shown that HbS frequency can greatly vary over short distances.³¹ We present the first evidence-based estimates with precision measures, accounting for these spatial heterogeneities (where data permits) and demographic factors. Our model-based predictions showed large differences at the national level when compared with previous estimates (appendix pp 29–33, 45–48).⁷ Most neonates with HbS are born within specific regions in Nigeria, DR Congo, and India (figures 3–5, appendix pp 29–44). Health interventions (eg, screening programme, genetic counselling) in those regions would, therefore, have the greatest effect on total numbers of affected neonates.

Due to the kinetics of inherited disorders, the current and future public health challenges are twofold. First, countries with high HbS allele frequencies need to be prepared to provide health care for patients with sickle cell anaemia. With substantial reductions in mortality of children younger than 5 years of age in many low-income countries,⁸ SS neonates who would have previously died undiagnosed are now more likely to survive to adulthood. These countries, therefore, need to know the extent of the problem and to plan efficient and affordable treatment policies to reduce the morbidity of these patients during their lifetime.⁹ Second, although AS individuals are usually asymptomatic, it is important to educate carriers about their status to try to prevent the future birth of homozygous children, as well as to avoid psychosocial issues (eg, misconceptions or stigmatisation).³² Such education programmes are needed not only in low-income countries but also in high-income countries.

Without a universal neonate screening programme and effective data aggregation and analysis systems (appendix pp 25–26), it is impossible to know with exact precision the number of neonates affected by HbS (or any other inherited disorder) within a given country. Pilot screening programmes are underway in several African countries and India, which should help update our knowledge of the current distribution of HbS and refine our estimates in many areas. Better accessibility to data in high-income countries (eg, USA and UK) would also be valuable.

While the approach presented in our report represents a substantial advance in both the methodology and underlying evidence base compared with earlier attempts to quantify global estimates of neonates affected by sickle haemoglobin,⁷ several limitations remain. These limitations include the scarcity or accessibility of information about HbS allele frequency in some important areas; the presence of confounding factors such as non-random mating; and the interactions with other inherited blood disorders.

Contemporary HbS allele frequency heterogeneity over short distances and between ethnic groups can only partly be explained by the selection pressure of malaria, especially in India.⁴ Other factors, such as environmental variables (eg, altitude) or differences in social behaviour (eg, consanguinity or endogamy)³³ also contribute to HbS allele frequency. For various reasons, including ethical issues and the difficulty of defining an ethnic group, such social factors are not consistently reported in HbS surveys. Therefore, capturing spatial variations in these factors from a partial dataset is difficult. As a consequence, the inclusion of ethnic origin or consanguinity would probably have added uncertainty to the model. Moreover, no consistent high-resolution global database of these factors is currently available. Defining a minimum standard of information gathered and reported for new surveys (eg, sampling method, location, age group) would greatly facilitate such global studies.

Finally, our study focuses on AS and SS neonates. We are aware that these estimates only represent part of the burden caused by HbS. First, we have calculated neonate estimates, not all-age population estimates or disability-adjusted life year losses from HbS.¹ The feasibility of calculating such estimates is currently very limited because of a general paucity of information on the variations of clinical severity between haplotypes, and on mortality caused by HbS, both in low-income and high-income countries.³⁴ Second, our estimates do not include individuals with compound disorders such as haemoglobin SC disease or haemoglobin S-beta thalassaemia disease. Including these individuals would require a good knowledge of the distribution of haemoglobin C and β-thalassaemia, as well as of the interactions between these variants. Recent studies have shown that such interactions, or epistatic effects, are complex.^35,36 Unless similar rigorous quantitative methods are used to study related disorders and HbS mortality, reliable estimates of the general economic and public health burden of haemoglobinopathies as a whole will be impossible to produce.

Having been neglected for decades, awareness of HbS and of its burden on public health services globally is finally growing. Our study aimed to make optimal use of existing data to create a contemporary map of HbS allele frequency and calculate annual estimates of AS and SS neonates at various spatial scales. The global quantitative map presented here shows for the first time the contemporary distribution of HbS. Our estimates and associated precision measures provide a valuable result to guide public health decisions at various scales and to assess future changes.