The clinical epidemiology of sickle cell anemia In Africa

Abstract Sickle cell anemia (SCA) is the commonest severe monogenic disorders of humans. The disease has been highly characterized in high‐income countries but not in sub‐Saharan Africa where SCA is most prevalent. We conducted a retrospective cohort study of all children 0–13 years admitted from within a defined study area to Kilifi County Hospital in Kenya over a five‐year period. Children were genotyped for SCA retrospectively and incidence rates calculated with reference to population data. Overall, 576 of 18,873 (3.1%) admissions had SCA of whom the majority (399; 69.3%) were previously undiagnosed. The incidence of all‐cause hospital admission was 57.2/100 person years of observation (PYO; 95%CI 52.6–62.1) in children with SCA and 3.7/100 PYO (95%CI 3.7–3.8) in those without SCA (IRR 15.3; 95%CI 14.1–16.6). Rates were higher for the majority of syndromic diagnoses at all ages beyond the neonatal period, being especially high for severe anemia (hemoglobin <50 g/L; IRR 58.8; 95%CI 50.3–68.7), stroke (IRR 486; 95%CI 68.4–3,450), bacteremia (IRR 23.4; 95%CI 17.4–31.4), and for bone (IRR 607; 95%CI 284–1,300), and joint (IRR 80.9; 95%CI 18.1–362) infections. The use of an algorithm based on just five clinical features would have identified approximately half of all SCA cases among hospital‐admitted children with a number needed to test to identify each affected patient of only fourteen. Our study illustrates the clinical epidemiology of SCA in a malaria‐endemic environment without specific interventions. The targeted testing of hospital‐admitted children using the Kilifi Algorithm provides a pragmatic approach to early diagnosis in high‐prevalence countries where newborn screening is unavailable.


| I N TR ODU C TI ON
Sickle cell anemia (SCA) is a major global public health concern of which sub-Saharan Africa bears the greatest load. More than 3 out of 4 of all those affected worldwide are born within the region-almost a quarter of a million new births every year. 1 The allele responsible, the b s mutation in HBB, is the textbook example of a balanced polymorphism in humans. 2 The sickle mutation has been selected to high population frequencies in many tropical regions because carriers (HbAS) are strongly protected against death from P. falciparum malaria 3 ; however, homozygotes with SCA suffer chronic ill health and reduced survival 1 and even today, without specific treatment most children born with SCA in sub-Saharan Africa will die before their fifth birthday. 4 Although the clinical epidemiology of SCA has been well described in resource-rich regions, most notably through the Cooperative Study of Sickle Cell Disease in the United States, 5 few studies have been conducted in sub-Saharan Africa. 6 This is important because the natural history of SCA is likely to differ in the African context for reasons that include the heavy burden of malaria and other infections, more limited access to medical care and high rates of undernutrition. 7,8 In this study, our aim was to document the common complications of SCA in a typical African hospital. Specifically, we wished to investigate the current and future burdens placed by SCA on African health facilities and the potential utility of targeted screening among specific subgroups of hospital-admitted children as an alternative to newborn screening in the community.

| Study population
The study was conducted on the pediatric wards of Kilifi County Hospital on the Coast of Kenya. This serves as the first referral hospital for Kilifi County where the common causes of admission are similar to those in many hospitals in sub-Saharan Africa. 9 In 2000, the Kilifi Health and Demographic Surveillance System was established in a defined area of 891 km 2 surrounding Kilifi County Hospital, which includes a population of around 100,000 children younger than 14 years. 10 Approximately 80% of children who are admitted to Kilifi County Hospital reside within the Kilifi Health and Demographic Surveillance System study area. 10 At the time of sampling, and even to date, there was no routine screening for SCA anywhere in Kenya nor any specific nationally based guidelines regarding management.
Hydroxyurea treatment, now commonly used in high-income countries, was not available.

| Study participants
A system of routine clinical surveillance has been operating at Kilifi County Hospital since 1989 through which trained clinicians assess all children at both admission and discharge, and record standard data on a computerized proforma. 10 A range of laboratory tests are conducted on all admitted children which include a full hemogram, a malaria blood film, and a blood culture. 11 Admission samples from this surveillance system have been archived at 2808C since 2000. This study included all children younger than 14 years who were residents of the Kilifi Health and Demographic Surveillance System study area and who were admitted to Kilifi County Hospital between January 2000 and December 2004. We chose this study interval because during this period malaria was the commonest cause of pediatric admission 9 and the incidence of uncomplicated malaria in the surrounding community was between one and three episodes/child/year, 9 allowing us to address the important question of the relative contribution of malaria to ill health in children with SCA in Africa. Furthermore, this predated the introduction of routine immunization against pneumococcal infections and of the wide-spread diagnosis and provision of care for children with SCA. As such, the study is broadly representative of the natural history of SCA in the absence of specific interventions. All cases were classified according to their SCA status (SCA -HbSS; non-SCA -HbAA or HbAS), determined through the genotyping of archived admission samples during 2015 and 2016, and also on the basis of a range of clinical and laboratory features (Table 2). Finally, through careful inspection of their clinical notes, confirmed cases of SCA were further classified according to a number of additional features that were specific to children with SCA but that are not ordinarily captured by our clinical surveillance database. Such complications included pain, hand-foot syndrome and priapism, as defined in Table 5. Incidence rates were computed through use of the mid-study population of the Kilifi Health and Demographic Surveillance System study area 10 stratified by the age-specific prevalence of SCA. As described in detail previously, we determined the latter by typing for SCA children who we recruited by random sampling throughout the Kilifi Health and Demographic Surveillance System study area into studies undertaken between 1998 and 2005. 12

| Laboratory procedures
Admission hemograms, blood film examinations for malaria, and blood cultures were performed as previously described. 9,11 We tested cases for SCA by PCR 13
Continuous data were compared using parametric or nonparametric tests as appropriate, while proportions were compared using the v 2 test. Sensitivities and positive predictive values (PPVs) were computed using the "diagt" command. We calculated the incidence of syndromespecific admission to Kilifi County Hospital in SCA and non-SCA children from the mid-study population of the Kilifi Health and Demographic Surveillance System area and the age-specific prevalence of SCA among controls. 12

| RE S U L TS
A total of 20,574 children <14 years were admitted to the wards of Kilifi County Hospital from within the Kilifi Health and Demographic Surveillance System study area during the study period, of whom 18,873 (92%) were genotyped successfully and included in this analysis (Supporting Information Figure). The demographic, anthropometric and hematological characteristics of children at the point of admission, stratified by SCA status, are summarized in Table 1. We found no evidence for a prior diagnosis in 399 of 576 (69.3%) SCA admissions.
Overall, children with SCA accounted for 3.1% of all admissions and were typically older, less-well nourished, and differed from children without SCA across a range of hematological parameters (  Information Table S1). Compared with non-SCA admissions (7610/18292; 42%), those with SCA were less likely to be positive for P. falciparum malaria parasites (98/576; 17%; P < .0005; Table 2) and parasite densities were significantly lower in those who were infected (Table 1). While mortality among parasite-positive patients with SCA (4/98; 4.1%) was higher than that in those without SCA (199/7610; 2.6%), this did not reach statistical significance (v 2 5 0.81).
The clinical phenotypes and outcome of hospital admission, together with data on the potential value of such phenotypes as predictors of SCA, are summarized in Table 2. While children with SCA were over-represented among some clinically defined subgroups, including those with clinically detectable jaundice, severe anemia (Hb < 50 g/L) and a range of specific bacterial infections, they were under-represented among both children admitted with malaria and those admitted during the neonatal period. Although some syndromes, including jaundice, severe anemia and stroke, were associated with PPVs of >10%, the same diagnoses did not generally provide a sensitive basis for the targeted screening of hospital-admitted children. We therefore investigated whether a multi-disease approach to screening might provide a more efficient strategy. To answer this question we developed an algorithm based on a range of clinical syndromes ordered hierarchically on the basis of their sensitivity and PPV profiles. Given that, in general, it is unnecessary to screen children more than once, we limited this analysis to index admissions for SCA-patients admitted more than once. We re-classified each syndrome in this hierarchy to be mutually exclusive of the preceding syndrome, allowing us to estimate the cumulative PPVs and numbers needed to test (NNT) to identify each affected case (Table 3). From this analysis we conclude that by testing all children presenting with just one of five syndromes-stroke, severe anemia, jaundice, septic arthritis and cellulitis, pyomyositis, or abscess we would have identified 47% of all those with SCA by testing only 14% of all those admitted: a NNT to identify each affected child of 14. Testing all admitted children would have been more sensitive (100%) but at the expense of a NNT of 47.
The incidence rate ratios (IRRs) for admission with various syndromes in SCA compared to non-SCA children are shown in Table 4 while the equivalent incidence rates are summarized in Supporting Information Table S2. The overall incidence rates for all-cause admission were 57.2/100 person years of observation (PYO) (95% CI 52.6-62.1) and 3.7/100 PYO (95% CI 3.7-3.8) in SCA and non-SCA children respectively (IRR 15.3; 95% CI 14.1-16.6). However, in contrast to non-SCA children, the incidence rates for admission with a range of syndromes rose strongly with age among those with SCA and, as a result, we present these data stratified by age category (Table 4   common among SCA children with than without anemia (24.2% vs. 14.1%, respectively; P 5 .003), blood films were negative in the majority of anemic cases. Finally, the incidence rates for a range of specific complications of SCA are summarized in where historically, the disease has been widely neglected. 14   Clinical phenotypes were defined as follows: "neonatal conditions"-admission to hospital within the first 28 days of life; "malaria"-a fever in the presence of P. falciparum parasitemia at any density in children <1 year old or at a density of >2500 parasites/lL in older children; "severe malaria"-malaria in association with the specific complications of prostration, coma, respiratory distress, or a Hb of <50 g/L; "severe" and "very severe pneumonia"-defined as previously described 31 ; meningitis/encephalitis-the clinical detection of "neck stiffness," prostration or coma (defined as a Blantyre Coma Score of <5), or a bulging fontanel; "severe malnutrition"-a mid-upper-arm circumference of 7.5 cm in children <6 months or of 11.5 in children 6 months of age; "gastroenteritis"-diarrhoea (3 or more loose watery stools/day) with or without vomiting (3 or more episodes/day); "jaundice"-the clinical recognition of jaundice by the admitting clinician; "osteomyelitis"-bacterial bone infection diagnosed either clinically or with supportive radiological or microbiological evidence; "cellulitis/pyomyositis/abscess"-diagnosed clinically with or without supportive microbiological evidence; "septic arthritis" diagnosed clinically with or without supportive microbiological evidence; and "stroke"-sudden onset of unilateral weakness persisting for more than seven days and for which other diagnoses had been excluded. Some children contribute data to more than one row.   Table 2 using the "diagt" command in Stata v14.2. Syndromes were defined as described in Table 2. The IRRs for each syndrome and age group were estimated using a series of Poisson regression models. In each syndrome-specific model the count of that syndrome was the outcome variable and the indicator for SCA was the explanatory variable. within the study area, it is likely that some children will have sought care in hospitals beyond the study borders. Second, it is likely that the costs associated with medical consultation and hospital admission will have deterred some parents from seeking care. As a result, our study may not have captured all significant events among resident children with SCA.
In contrast to observations made in resource-rich countries through studies like the US cooperative study for sickle cell disease, 5 where hand-foot syndrome was a common presenting feature of SCA, its incidence was low in the current study. While this might simply reflect methodological issues, it could also result from a true difference in the clinical phenotype of SCA within our study population, potentially explained by genetic or environmental factors. Conversely, severe anemia was significantly more common in our study than in reports from the north. For example, while the overall incidence of anemia in the cooperative study was <5/100 PYO, 5  While the design of our study does not allow us to determine the precise etiology of these anemic episodes, the most likely explanations for this geographic discrepancy relate to malaria or nutrition. Although children with SCA are thought to be partially resistant to malaria infection, 17 a theory supported by the lower prevalence and density of P.
falciparum infections seen in this study, the disease can take a particularly fulminant course when patients are infected. 18   PYO were derived as described in Table 4. Specific features were determined through physical inspection of the clinical notes of all children with SCA. Pain, any mention of un-explained pain; hand-foot syndrome, painful swelling of hands or feet; arthralgia, specific mention of pain in any joint. Some patients manifest more than one complication. No episodes of priapism were identified during the course of this study. Some patients manifest more than one complication. sodium metabisulphite sickling test represent potential approaches to screening, they both require a basic laboratory staffed by a trained technician and lack both specificity and sensitivity with regard to differentiating trait from disease. However, the recent development of simple and accurate point-of-care testing devices for SCA 29 now makes opportunistic testing at the bed-side much more practical. We recommend that where newborn screening is not available, hospitals in Africa should use the Kilifi Algorithm coupled to point-of-care testing as a pragmatic alternative.
In summary, we have described the clinical epidemiology of SCA within a malaria-exposed population in Africa. We highlight important differences in comparison to descriptions from resource-rich regions including a significantly higher incidence of severe anemia. We show that targeted screening of hospital-admitted children based on a small number of easily recognized syndromes provides a practical and efficient way to identify a high proportion of affected children and introduce them to care. If the world continues to ignore the problem of SCA, the disease will continue to result in high child-mortality and threaten the ability of many countries in sub-Saharan Africa to meet their Sustainable Development Goals. 30 The need is urgent for a renewed and increased focus on SCA by both African governments and the international community.