Portal de Periódicos da CAPES

Artigo Acesso aberto Produção Nacional Revisado por pares

BIBTEX RIS

Estimating the post-neonatal prevalence of sickle cell disease in a Brazilian Population

2010; Wiley; Volume: 15; Issue: 10 Linguagem: Inglês

10.1111/j.1365-3156.2010.02597.x

1365-3156

Chinenye Ilozue, Rosana Cipolotti, Carlos André Araújo Melo, Ricardo Queiroz Gurgel, Luís E. Cuevas,

HIV, Drug Use, Sexual Risk

Objective To estimate the prevalence of individuals with sickle cell disease (SCD) in Aracaju, Brazil, using the capture–recapture (CRC) method. SCD is a significant public health problem with long-term life-threatening complications. There are no reliable estimates of the number of individuals with this condition in Aracaju, north-east Brazil. The CRC method has been used to quantify other ubiquitous populations. Method Three independent lists of individuals with homozygous (HbSS) SCD were constructed from patients attending the main specialist ambulatory service, all patients with SCD admitted to three government hospitals and a clinic providing specialist immunisation services to patients with SCD. Individuals were matched to ascertain whether they appeared in one, two or three lists, and population size was estimated using the log-linear model. Results The lists identified 374 individuals. Two hundred and one appeared in one, 99 in two and 74 in three lists with an estimated number 400 (95% CI 387–418) HbSS SCD individuals; 51.6% patients with SCD were men and age ranged from 1–62 years (median 14). Conclusion The CRC method resulted in a smaller population estimate than expected. The causes of this discrepancy may include list dependence, high mortality with a survival cohort effect and the method of identifying the more severe cases. The CRC method has potential to estimate the size of this population and could supplement neonatal screening to further characterise the SCD population in this region. Estimation de la prévalence post-néonatale de la drépanocytose dans une population brésilienne Objectif: Estimer la prévalence des personnes atteintes de drépanocytose à Aracaju, au Brésil, en utilisant la méthode de capture-recapture. La drépanocytose est un problème de santé publique important avec des complications vitales à long terme. Il n’existe aucune estimation fiable du nombre de personnes avec cette condition à Aracaju, dans le nord du Brésil. La méthode de capture-recapture (CRC) a été utilisée pour la quantification dans d’autres populations. Méthode: Trois listes indépendantes d’individus avec une drépanocytose homozygote (HbSS) ont été construites à partir des patients fréquentant le principal service ambulatoire spécialisé, tous les patients atteints de drépanocytose admis dans 3 hôpitaux gouvernementaux et une clinique offrant des services de vaccination spécialisée pour les patients atteints de drépanocytose. Les individus ont été comparés pour savoir s’ils apparaissent dans un, deux ou trois listes et la taille de la population a été estimée au moyen du modèle log-linéaire. Résultats: Les listes ont identifié 374 individus. 201 ont paru dans une, 99 dans deux et 74 dans trois listes avec un nombre estiméà 400 individus (IC95%: 387-418) atteints de drépanocytose HbSS. 51,6% des patients atteints de drépanocytose étaient de sexe masculin et l’âge variait de 1 à 62 ans (médiane 14). Conclusion: La méthode de CRC a résulté en une estimation plus faible que prévue de la population. Les causes de cet écart peuvent comprendre la dépendance des listes, une forte mortalité avec un effet de cohorte de survie et la méthode d’identification des cas les plus graves. La méthode de CRC a le potentiel pour estimer la taille de cette population et pourrait compléter le dépistage néonatal afin de mieux caractériser la population des drépanocytaires dans cette région. Estimando la prevalencia postnatal de la Enfermedad de Células Falciformes en una población brasilera Objetivo: Estimar la prevalencia de individuos con la Enfermedad de Células Falciformes (ECF) en Aracaju, Brasil, utilizando el método de Captura y Recaptura. La ECF es un problema importante de salud pública, con complicaciones a largo plazo y riesgo de muerte. No existen estimaciones fiables del número de individuos con esta condición en Aracaju, noreste de Brasil. El método de Captura y Recaptura (CRC) se ha utilizado para cuantificar otras poblaciones ubicuas. Método: Se construyeron tres listas independientes de individuos homocigotos para (HbSS) de pacientes que visitaban al principal especialista del servicio ambulatorio, todos los pacientes con ECF admitidos a los 3 hospitales gubernamentales y a una clínica que prestaban servicios especializados de inmunización a pacientes con ECF. Los individuos fueron pareados para establecer si aparecían en una, dos o tres listas y se estimó el tamaño poblacional utilizando un modelo logarítmico linear. Resultados: Las listas identificaron a 374 individuos. 201 aparecían en una, 99 en dos y 74 en tres listas con un número estimado de 400 (95%IC 387-418) individuos homocigotos para ECF. 51.6% pacientes con ECF eran hombres, con un rango de edad de 1-62 años (mediana 14). Conclusión: El método CRC resultó en una estimación de la población menor que la esperada. Las causas de esta discrepancia podrían incluir: dependencia de la lista, alta mortalidad con un efecto de cohorte de supervivientes y el método para identificar los casos más severos. El método CRC tiene el potencial para estimar el tamaño de esta población y podría suplementar el tamizaje neonatal para caracterizar más a fondo la población con ECF en esta región. Sickle cell disease (SCD) is the most common genetic haematological disorder globally. It is an inherited disorder of haemoglobin (Hb) that leads to the pathological haemolysis and vaso-occlusion (Hoffbrand et al. 2005; Booth et al. 2009) resulting in a wide variety of clinical features (Loureiro & Rozenfeld 2005b) and high mortality, particularly in young children. Many national health services screen newborns for the early detection of SCD (Tshilolo et al. 2008), and in Brazil, neonatal screening began in Sao Paulo in 1992. Although the programme has since expanded (Brandelise et al. 2004), it has not been implemented countrywide, particularly in the north-east, where the prevalence of SCD is high (Bandeira et al. 2008). Since 1991, Brazil adopted a national health policy whereby the whole population has the right to access health services and created the Sistema Único de Saúde (SUS, National Health System) and its community extension, the Family Health Programme (Programa de Saúde da Família- PSF). There also exist other programmes to address the socio-economic and healthcare needs of Brazil’s black population. These programmes have increased the population’s access to health services and decreased mortality over the last decades (Menicucci 2009). Although the survival of individuals with SCD seems to have improved in Brazil in recent decades, there are no reliable statistics on the prevalence of SCD after birth or centralised registers to quantify the burden of SCD in the country. Patients with sickle cell disease are ubiquitous and mobile and less conventional methods than routine surveillance are necessary to describe their characteristics and numbers. One of these potential methods is the capture–recapture (CRC) method, which was initially developed to quantify animal and insect populations, and has been applied to describe difficult to reach human populations. Examples of its use include patients with type 1 and 2 diabetes, intravenous drug users, sex-workers and street children (Nanan & White 1997; Gurgel et al. 2004). The method uses lists of individuals constructed from patients attending several locations and estimates the population size by identifying the number of individuals repeatedly appearing in several of the lists (Hook & Regal 1995; IWGDMF 1995b, Nanan & White 1997; Chao et al. 2001). Although the method seems suitable to quantify SCD, as patients are ubiquitous, there are no reports of its use in this population. Knowledge of the prevalence of SCD and its morbidity and mortality burden on health and social services would enable effective health care planning to address their needs. This study therefore describes the use of the CRC method to assess the prevalence of SCD in Aracaju, Sergipe State, in north-east Brazil, and the characteristics of the patients identified by the method. The study was conducted in Aracaju City, Sergipe State, in north-east Brazil (Figure 3). Aracaju and Sergipe have an estimated population of 536 785 and 1 999 374 (2008) inhabitants, respectively (IGBE 2009). Patients with known HbSS status resident in Sergipe were eligible for inclusion into the study, and patients with other known haemoglobinopathies were excluded. Three lists of individuals with SCD were constructed to apply the capture–recapture method. Geographical procedence of patients with sickle cell disease in Sergipe State. One list consisted of all patients with known SCD attending the paediatric and adult haematology outpatients’ clinics at the University Hospital of Sergipe’s Federal University. Records for all SCD individuals were identified by manually reviewing the clinic’s records and computerised databases since the opening of the clinic in 1987. The second list was constructed with all patients with a diagnosis of SCD admitted to the three public hospitals of Aracaju from 2006 to 2008. The list was constructed by requesting all computerised databases of the hospital admissions and identifying patients who had a diagnosis of SCD from their medical records (clinical history and haemoglobin electrophoresis), independent of the cause of admission. All admissions and relevant clinical records were reviewed to ensure no HbSS individuals were missed. The third list was constructed with known patients with SCD registered at the Centro de Referência em Imunobiológicos Especiais (Reference Centre for Special Immunobiologics, CRIE). The CRIE provides special immunisation services with vaccines not routinely included in the National Immunisation Programme to high-risk populations. Data collected from these sources included the patients’ name (forenames and surnames), date of birth and gender. These data were used to manually match individuals within and across the lists. Address and age were collected but not used because this information is changeable and was often incomplete in the registers. Based on an estimated incidence of 0.3% newborn infants having SCD, it was expected that Aracaju would have about 1600 individuals with SCD. It was also expected that the prevalence of SCD in this region is higher than the national average because of the higher prevalence of black and mixed-race individuals in the region. After the collection of data and elimination of duplicate entrances in a single list, the lists were matched to identify individuals who appeared once, twice or three times. The appearance of individuals in the clinic (A), admissions (B) and CRIE (C) lists resulted in seven potential permutations (appearance in either A, AB, AC, B, BC, C or ABC). After matching and delineating the numbers of overlaps, the ‘recap’ subroutine for Stata 9.2 was used to run a log-linear regression model for capture–recapture analysis using the command ‘recap – frequency a b c’ which runs a log-linear model on the frequencies appearing in the overlaps between the lists. The use of ‘information criteria’ enables the selection of the most appropriate model for the population and associates the model with its likelihood ratio statistic. The Akaike information criterion (AIC) or Bayesian information criterion (BIC) (IWGDMF 1995a) enables the investigator to select the most appropriate model for the population and is the most widely adopted. The AIC is derived from the formula: AIC = −2 × [log(L) − q] where ‘log(L)’ is the log-likelihood as ascertained by the maximum likelihood estimates of the components of the model and ‘q’ represents the number of the parameters or components of the model (IWGDMF 1995a, Chao et al. 2001). The model that provided the smallest AIC value was selected to appropriately apply the AIC criterion. The analysis produced an estimate of the population size, confidence intervals and Akaike and Bayesian information criteria. This estimate is for patients attending health services in Aracaju, the capital city of Sergipe State. As lists include individuals’ resident in other areas of the state because patients from rural areas are often referred to the capital city, the estimate is of the prevalence of patients with SCD seen in Aracaju (therefore, the number of patients treated and followed up in Aracaju city). This study was approved by the Ethics Committees of the Federal University of Sergipe and the Liverpool School of Tropical Medicine. A total of 374 individuals appeared in at least one of the lists, whose demographics are described in Table 1. There was a predominance of men (51.6%) and a mean age of 17.1 years old and median age of 14. The ages of the SCD population ranged from 1 to 62 years with a median of 14 years (Figure 1). Age distribution of patients with sickle cell disease, Aracaju, Sergipe, Brazil. The three lists comprised 119 patients on list A, 151 on list B and 351 on list C. Seventy-four individuals appeared on the three lists and the number of individuals appearing in each or multiple lists is shown in Figure 2. Venn diagram of the number of patients identified by each study list (A = University Hospital outpatient SC clinic, B = Hospital admission, C = special vaccination service). The total population of HbSS SCD individuals derived from the log-linear analysis was 400. The AIC for this estimate was −2.12, and the model estimated that there were 26 missing cases (Table 2). Thus, the estimated prevalence was close to the manual count of 374 individuals. Sickle cell disease largely affects populations of Africa and is more widely prevalent in Brazil among the black and mixed-race descendents of Africans following the transatlantic slave trade. (Bandeira et al. 2008) Despite a historical neglect of diseases of indigent populations, Brazil has implemented a series of programmes to affirmatively address the socio-economic and healthcare needs of the black population. These initiatives, such as Seppir (Secretaria Especial de Políticas de Promoção da Igualdade Racial– Special Secretariat for the Promotion of Racial Equality) (Secretaria et al. 2008) and the Ministério da Integração Nacional (Ministry of National Integration), aim to minimise regional socio-economic disparities and address healthcare issues such as SCD in the black community (Ministry for National & Integration 2007). If successful, patients with chronic diseases, such as SCD, may experience improved survival in the next generations, and methods to quantify their numbers and monitor changes in their characteristics would be highly desirable. To date, the CRC method has not been used to estimate the population of SCD, and there is no data on the prevalence of SCD in north-east Brazil. The estimate of 400 cases derived through the CRC method is smaller than the estimate of 1610 (or 0.3% of 536 785 population), based on the prevalence obtained through neonatal screening (Bandeira et al. 2008). Thus, several possibilities need to be considered. It is possible that the SCD population in this area is truly small because of a lower prevalence of this area. This possibility, however, is very unlikely, as the north-east of Brazil had historical waves of migration of African and Caribbean populations with a high prevalence of SCD. In fact, it is generally accepted that the prevalence of SCD is higher in the north-east area than in the southern regions of the country. A further explanation may be high mortality owing to poor access to and low utilisation of health services with poor survival and migration of older generations. Even in a relatively major city such as Aracaju, with a reasonable public health system, there may be barriers to access to health care services for the poorer population. These could include lack and cost of transport (e.g. causing non-attendance to follow-up clinics), access to drugs and vaccines and lack of education and understanding of the disease, which are recognised barriers to the adequate long-term management of patients with chronic diseases. This possibility is supported by the age distribution of the cases included, with a peak age under 20 years, rapidly decreasing numbers in adulthood and very few patients being older than 30 years. It is also possible that this sample size is an underestimate because of the presence list dependence. List dependency is related to the choice of sources and is a recognised problem of CRC (IWGDMF 1995a) Individuals ‘captured’ in a list must have the same probability of being captured in other lists – thus, there is no positive dependence (or the appearance in one list increasing the chance of appearing in another) or the inverse (negative dependency). Positive dependency produces an underestimate, whereas negative dependency overestimates the true population size. Although the lists were constructed independently from databases of different institutions, the patients that access these services were likely to attend more than one centre as once individuals are included in a specialised health service; they are readily referred to the other specialist services. This would have resulted in a large degree of overlap between the lists and therefore positive dependence and an underestimate of population size. Furthermore, it is assumed that all patients have the same likelihood of experiencing severe clinical complications and thus attending emergency specialist health services, which is not the case for SCD, and that all patients with SCD are recorded when attending the hospitals, which is also unlikely to be the case, especially in patients with frequent hospital consultations. The heterozygous (HbAS – sickle cell trait) state is thought to have a national prevalence of 2–6% in Brazil and to be most frequently found in the south-east and the north-east, where the slave trade was practised most prominently (Bandeira et al. 2008, Loureiro & Rozenfeld 2005a; Salzano 1985). Bahia, Sergipe and Pernambuco in the north-east thus are thought to have a higher prevalence of SCD (Adorno et al. 2008), with the Bantu or Central African Republic (CAR) haplotype of the β-globin gene being the most prominent form. The ‘CAR’ haplotype is accepted to have a more severe clinical course than other haplotypes as the levels of foetal haemoglobin are lower (Bezerra et al. 2007; Adorno et al. 2008). The population derived here thus would represent only the most clinically severe individuals as these are the patients most likely to access health services. Heterozygous patients were not expected to be detected in any of the three lists. Estimates from recent mortality and morbidity patterns elsewhere estimate that approximately 50% of children with SCD born in the 21st century will live over 50 years of age (Powars et al. 2005), and the use of health services by surviving patients may change as current children who survive to older age have fewer episodes of crises (Mckerrell et al. 2004). In a 2004 study, only 25% of older patients (>40 years of age) had disease severity that warranted hydroxyurea therapy compared to 50% of younger patients with more severe disease (21–30 years of age), suggesting that those reaching older age have survived as a result of less severe disease, reflecting a cohort effect with attrition of patients with severe manifestations (Mckerrell et al. 2004). There were 186 patients in list C alone (older patients) and 188 appearing in the overlaps and the other lists, demonstrating that within our study population, approximately 50% of patients may represent a relatively milder disease course. Furthermore, a study of haemoglobinopathies among blood donors in Sergipe reported a prevalence of HbAS sickle cell trait of 4.1% (Vivas et al. 2006). Although the study population presenting to vaccination services may be biased because of better survival of individuals with milder disease presentation or socio-economic factors facilitating service accessibility, if a similar relationship can be assumed in this population, the 400 individuals derived here represent 50% or less of the actual total population with clinically severe (or moderate-severe) disease. Therefore, the true total population of SCD individuals would be in the order of ≥800 patients. Missing the milder cases of type II diabetes from all the lists/sources was an observation reported by Verlato and Muggeo (2000) in their use of CRC in the epidemiology of type 2 diabetes. They also reported similar list dependence and had a particularly heterogeneous population. Furthermore, a clinic-based study in Jamaica reported that patients with milder SCD are entirely missed by hospital databases as these individuals do not access care (Wierenga et al. 2001). The health care facilities from which the lists were constructed manage most patients with SCD in Aracaju and surrounding areas (Figure 3). There are, however, smaller regional health centres where patients may present for minor complaints and some patients may have been missed, as described for other conditions. As neonatal screening for haemoglobinopathies is introduced to more regions of Brazil, the CRC method has the potential to evaluate the completeness of the population screening and monitor changes in the SCD population structure after birth, thus facilitating the description of mortality patterns. List dependence thus is the main limitation of the estimate obtained. Although this violation of the CRC assumptions is to some extent bypassed by the log-linear model, the analysis does not eliminate all of the error caused by dependence. There is ongoing debate in the literature regarding the possible solutions to this problem (Hook & Regal 1995, IWGDMF 1995a). The application of CRC may be a useful tool for assessing the population size for SCD in Brazil. Although the population size obtained is likely to be an underestimate, it reveals areas for further study and provides an estimate of the number of patients with severe SCD utilising the health services of the State. The limitations of the method are largely related to list ascertainment and source selection, restricting the range of individuals included. Issues such as poor access to health care, milder (thus not formally acknowledged) clinical disease and dependence of the lists may all influence the precision of the estimate. The 400 cases identified most likely represent the more severe end of the disease spectrum, and an expansion of the origin of the lists may be necessary to ensure that milder (and probably older or not yet identified) cases are included to obtain a more comprehensive perspective of the problem. Neonatal screening is owing to be implemented in Sergipe in 2010. The improved documentation of a cohort of children with SCD will provide a unique opportunity to monitor the morbidity patterns of future generations and whether improvements in resources, access to services and better quality of data result in improved survival. We thank Professora Vera França from the Dept. of Geography, Federal University of Sergipe who produced the maps of Sergipe municipalities and the staff at CRIE who provided the third list. We are grateful to Dr B. Faragher, Head Medical Statistician at LSTM, for statistical advice and to Dr C. Baum at Boston College for input with the software.