Estimated mortality rates of individuals with sickle cell disease in Brazil: real-world evidence

Sickle cell disease (SCD) is a group of inherited chronic multisystem disorders, including sickle cell anemia, caused by mutations in the gene expressing β-globin.¹ SCD is a common and life-threatening inherited hemoglobin disorder, affecting ∼276 000 infants each year globally.² The global metaestimate for the birth prevalence of homozygous SCD is 112 per 100 000 live births, with a birth prevalence of 1125 per 100 000 in Africa compared with 43 per 100 000 in Europe.³ 

In the last decades, there have been changes in the clinical course of SCD, including an increase in diagnosis and decrease in mortality rates.⁴ Nevertheless, the impact of the disease is still significant, with affected individuals showing reduced quality of life and survival. Patients with SCD experience worse health-related quality of life than the general population, and their scores are similar to those of patients who undergo hemodialysis.⁵ In the context of the United States, a modeling study showed lower projected life expectancy (54 vs 76 years) and lower quality-adjusted life expectancy (33 vs 67 years) for individuals with SCD compared with a matched non-SCD cohort.⁶ 

In Brazil, SCD is one of the most common hereditary diseases.⁷ The Brazilian Ministry of Health estimates that between 60 000 and 100 000 individuals have SCD.⁷ From 2015 to 2019, the incidence was 45.92 cases per 100 000 live births, based on data from the National Newborn Screening Program.⁸ In previous studies, estimated mortality rates have ranged from 0.115 per 100 000 to 0.54 per 100 000 individuals.⁹ Main causes of death are infection, acute splenic sequestration, sepsis, and acute chest syndrome.⁹ Nevertheless, studies with national data are sparse, and most published results include data from a single state.^9-16 It is also essential to note that incidence and prevalence of SCD varies substantially among the regions in Brazil, as does access to health care.^1,17,18 

Furthermore, the disease burden of SCD is high in Brazil.^9,19 Several factors may negatively affect the health-related quality of life of these patients, such as the intensity and frequency of painful crises (caused by vaso-occlusion due to the sickling of red blood cells), the need for hospitalization and blood transfusion, and social circumstances (unemployment, low educational level, and financial concerns).²⁰ Moreover, observed rates of depression, anxiety, and even alcohol abuse are high.^21-24 In Brazil, these problems are magnified because of suboptimal health care system organization and barriers to adequate health care access for many patients with SCD.²⁵ 

Considering the paucity of epidemiological data on SCD mortality in Brazil, the substantial burden of the disease, and the need to improve care for individuals with SCD, the objective of this study was to provide mortality-related estimates to inform health care decision-makers for policy interventions. Specifically, we aimed to estimate the mortality risk for individuals with SCD compared with the general population using real-world data from Brazilian public databases. We also aimed to explore the trends in mortality over the last 5 years and the median estimated survival of individuals with SCD based on sex, race, and region of residence.

This study consists of a real-world evaluation of SCD mortality patterns in Brazil. Using data from public databases, we compared the median age at death reported in registries of individuals with or without SCD. We also estimated the crude mortality rate of individuals with SCD (per 100 000 inhabitants) and modeled survival curves for the general population and individuals with SCD, estimating and comparing the risk of death for an individual with SCD vs that for the general population and the mortality rate for both groups. In this article, we describe the methodology used in more detail.

We retrieved data from all registered deaths in Brazil from the Brazilian public database of the Mortality Information System (SIM [Sistema de Informações sobre Mortalidade]). In Brazil, a physician must issue a death declaration for every death, and this document is the source of information for the SIM database. All deaths must be registered in the SIM, but it is estimated that 98% of deaths in Brazil are recorded and can be found in the SIM database.²⁶ This database provides microdata from death certificates, including information about the age, sex, and residence of the deceased, along with information on the cause of death.²⁷ We acquired data considering a 5-year period from 2015 to 2019. The data download was done in April 2021. We did not include registries from 2020 onwards because they were not completely available at the time of data acquisition and analyses.

For data manipulation, we identified the variables of interest based on the dictionary of variables from the SIM and removed unnecessary columns from our analyses. The columns that were retained for each record are presented in the supplemental Material 1 (supplemental Table 1).

Next, to identify the records of individuals with SCD, we selected records with an International Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) code referring to SCD in any field related to the cause of death. The subcodes included in our analyses were D57.0 (sickle cell anemia with crisis), D57.1 (sickle cell anemia without crisis), D57.2 (double heterozygous sickling disorders), and D57.8 (other sickle cell disorders). We excluded registries with the subcode D57.3 (sickle cell trait) because this subcode only indicates that the individual carried 1 copy of the sickle β-globin gene, and it is not considered a disease. Although is possible for a patient with SCD to be wrongly registered under the code D57.3 at the time of their death, we understand that this possible misclassification bias has little impact on our findings because only 41 records of 6 553 132 records evaluated, presented this ICD-10 subcode. If the record presented the code D57 (sickle cell disorders) without a subcode, we include it in our analyses. We understand there is only a slight chance this would refer to a patient with sickle cell trait because the D57.3 subcode is frequently not reported in death certificates. Of note, in the ICD-10 classification used in the Brazilian database, the subcode D57.4 (sickle cell thalassemia) is not used. We expected that patients with this condition were classified under 1 of the other subcodes, specifically D57.2 (double heterozygous sickling disorders) or D57.8 (other sickle cell disorders), and, therefore, were included in our analyses.

We used descriptive statistics to describe the sex, race, age at death, and cause of death recorded in the registries. Of note, in the databases used for this study, race/skin color is usually self-reported by the patients or reported by their relatives, based on 5 options developed by the Brazilian Institute of Geography and Statistics: branca (white), preta (black), parda (brown), amarela (yellow), and indígena (indigenous person).

We calculated the median age at death for both the SCD and general population, stratified by sex, race, and region of residence. For the overall estimation, we a conducted sensitivity analysis considering only registries with SCD as the primary (underlying) cause of death.

We also identified the main causes of death for SCD individuals, stratifying these data per age groups. Hematologists adjudicated the underlying cause of death.

Moreover, we estimated the crude mortality rate of individuals with SCD (per 100 000 inhabitants) and the general population for each year analyzed.

We modeled survival curves for the general population and individuals with SCD, considering the following parameters: the incidence of SCD in Brazil as 45.92 cases per 100 000 live births, as shown by National Newborn Screening Program data from 2015 to 2019, with an average coverage of 82.5%⁸; and a subnotification of SCD as the cause of death of 46% (ie, in 46 of 100 deaths of individuals with SCD, there is no mention of a related ICD-10 code on their death certificate).²⁸ For population size, we used estimates from the Brazilian Institute of Geography and Statistics for the same period.²⁹ For each age, we estimated the risk of death for an individual with SCD compared with the general population and the mortality rate for both groups. Data were censored at 90 years of age. The complete calculation log is included in the supplemental Material 2.

From these curves, we also estimated the median survival of the 2 groups, individuals with or without SCD. The median survival is the length of time from birth that half the patients of the group of interest are still alive. We highlight that the number of individuals with SCD in any given year is not directly measured, and the data used for analysis consist of estimated data, considering factors such as the incidence rate assessed through newborn screening and estimates of underdiagnosis of the disease. In contrast, the median age at death is directly observed from the registries, representing the midpoint of the frequency of the distribution of ages at death. Therefore, these estimates represent different and complementary information.

Data are presented as median (interquartile range [IQR]) or absolute number of individuals and percentages, unless otherwise stated. All analyses were conducted with R (version 4.1.0), packages Microdatasus (version 0.3.0), dplyr (version 1.0.5), stringr (version 1.4.0), ggplot2 (version 3.3.2), and RecordLinkage (version 0.4-12.1) and Microsoft Excel.

Between 2015 and 2019, a total of 6 553 132 deaths were registered in Brazil. From these records, we identified 3320 (0.05%) with an ICD-10 code or subcode of interest.

Among individuals with SCD, the median age at death was 32.0 years (IQR, 19.0-46.0 years; Figure 1A; Table 1), 37 years younger than that of the general population (median, 69.0 years; IQR, 53.0-81.0 years, Figure 1B; Table 1). Most were Brown or Black (78.6%), and 52.2% were women. In the general population, 44.9% of registries were of Brown or Black individuals, and 44.1% were of women (Table 1). Most deaths were observed in the Southeast and Northeast regions (44.9% and 34.6%, respectively), where the SCD genotype is more prevalent.^11,18 Of note, we did not observe a substantial difference between the place of birth and the place of residence at the time of death among the 3320 patients with SCD: only 310 (9.3%) were living in a different region at the time of death. This proportion was similar to that observed among the general population (n = 810 282; 13.5%).

Results were consistent when stratified by sex or race. For women, the median age at death was 34.0 years (IQR, 21.0-48.0) among individuals with SCD and 74.0 years (IQR, 60.0-85.0) among the general population. For men, it was 30.0 years (IQR, 18.0-43.0) and 66.0 years (IQR 48.0-78.0) for individuals with SCD and the general population, respectively. Among Brown or Black individuals, the median age at death was 31.0 years (IQR, 19.0-45.0) in the population with SCD and 65.0 years (IQR, 46.0-78.0) in the general population; among other races, the estimates were 35.0 years (IQR, 20.0-54.0) and 73.0 years (IQR, 59.0-84.0) for these same groups, respectively. We observed important differences in the median age of death between regions, ranging from 23.5 years in the North to 37.0 years in the South (supplemental Material 1 [supplemental Table 2]).

Results were consistent in the sensitivity analysis: considering only records with SCD as the primary cause of death, we identified 2603 records, and the median age at death was 30.0 years (IQR, 19.0-45.0).

Over the 5 years evaluated, the mean crude annual mortality rate of the population with SCD was 0.32 per 100 000 inhabitants of the general population; considering only deaths with SCD as a primary cause, the mean mortality rate was 0.25 per 100 000 inhabitants. In the general population, the mean crude annual mortality rate was 631.7 per 100 000 inhabitants. We did not observe substantial changes or statistical trends in the yearly mortality rate over the period evaluated (0.30-0.34 per 100 000 inhabitants; supplemental Material 1 [supplemental Figure 1]). The highest death rates were observed in the Center-West region (0.43 per 100 000 inhabitants), followed by that in the Northeast (0.41 deaths per 100 000 inhabitants) and Southeast (0.35 deaths per 100 000 inhabitants). The North and South regions presented the lowest rates, 0.22 and 0.10 per 100 000 inhabitants, respectively. Mean 5-year estimates per regions and states are shown in supplemental Material 1 (supplemental Table 3).

The most common causes of death among individuals with SCD-related mortality were sepsis (24.2%) and respiratory failure (7.9%). Similar results were observed for all age groups (supplemental Material 1 [supplemental Table 4]).

We estimate a prevalence of 60 017 individuals living with SCD, corresponding to 29.02 cases per 100 000 people. Moreover, the estimate for incidence was 1362 new cases yearly, on average, from 2015 to 2019. Estimated median survivals were 40 and 80 years for individuals with and without SCD, respectively. Modeled survival curves are presented in Figure 2. Mortality risk (Tables 2 and 3; Figure 3) was similar during the first year of age (relative risk [RR], 1.36); however, during childhood, the risk of death of individuals with SCD was >30 times higher in comparison with individuals without SCD (RR, 32.78). Between the ages of 20 and 39 years, the risk of death was 13 times higher in individuals with SCD (20-29 years: RR, 13.47 and 30-39 years: RR, 13.84). In individuals aged >70 years, the risk of death was similar between groups. Over the lifetime, SCD was associated with a loss of 32 years.

Our results showed that, in Brazil, SCD is associated with a reduction of ∼37 years of median age at death, from 69 years in the general population to 32 years in the population with SCD. Based on modeled data, the projected median survival among patients with and without SCD was 40 and 80 years, respectively. Moreover, SCD was associated with 32 years of life loss. The modeled mortality risk among children is >30 times higher than that of the general population; furthermore, young adults with SCD have mortality rates ∼13 times higher than those for individuals without SCD in the same age group. In older adults, the magnitude of risk observed is lower; in individuals aged >70 years, the risk of death is similar to that of the general population.

We observed substantial differences in the median age of death of individuals with SCD across regions; however, differences in the median age at death were also identified for the general population. Therefore, with the available data, it is not possible to determine whether the differences observed in the median age at death of the population with SCD are because of differences in health care across regions (ie, individuals in certain regions have poorer access to health care) or specific health care differences for the population with SCD (ie, there is a larger gap of health care access for individuals with SCD than for other individuals). Nevertheless, these results show that, despite the existing public health system with national coverage, there are important differences in the median age at death across regions, highlighting inequities in health care access in Brazil, which are clearly observed in the population with SCD.

As expected, the leading causes of death were septicemias and respiratory failure.^30,31 Similar causes were observed across all age groups. Importantly, most of these events could have been prevented with evidence-based therapies available in Brazil, such as antibiotics, vaccines, and hydroxyurea. In Brazil, there are published lines of care for SCD; however, there is a need for better implementation of their recommendations.³² 

Data from the literature reinforce our findings; the high impact of SCD mortality in Brazil is well known, and some previous studies have shown some concerning results.⁹ Of the 912 newborns with SCD in Rio de Janeiro from 2000 to 2010, 34 (3.7%) died, mainly because of acute chest syndrome (36.8%), sepsis (31.6%), or acute splenic sequestration (21.1%).¹² A cohort of 104 pregnant women with SCD in the state of Minas Gerais showed that one-third of them had near misses, and 4.8% died.³³ In Bahia, 1 of the Brazilian states with the highest prevalence of SCD, a study reported 74 deaths in 2011; ∼42% of the deaths occurred in adults aged 20 to 39 years.³⁴ Lobo et al published a comprehensive analysis of mortality in patients with SCD in Rio de Janeiro, Brazil, with data from 1676 patients, followed-up for over 15 years.³⁵ They observed an overall mortality rate of 16.77%; the mortality was higher in children aged between 6 and 11 years (35.9%). The main causes of death were infections (29%) and acute chest syndrome (25%). In this study, they estimated life expectancy at birth to be 53.3 years (compared with 74.6 years in the general population). Of note, this study was restricted to patients assisted in a single reference center; thus, a more critical impact of SCD is expected in other contexts in Brazil, as shown in our analysis.³⁵ 

The important impact of SCD on mortality has been observed worldwide. For example, in 2016, the overall life expectancy for Black individuals in the United States was 75 years; the average age of death among SCD-related deaths was 40 years.³⁶ In a modeled US cohort, the life expectancy of individuals with SCD was estimated to be 22 years lower than that of the general population (54 vs 76 years, respectively).⁶ In Jamaica, the median survival of individuals with SCD was estimated to be 53 years for men and 58.5 years for women.³⁷ These estimates are higher than what we observed in our study and may represent differences in health care access in these countries.

Our study has some limitations. Possible misdiagnoses and inaccurate completion of death certificates are potential limitations of our study. For instance, there was a decrease in the number of deaths registered at the Center-West region in 2017, which could indicate a gap in death ascertainment. Considering challenges related to accuracy of SCD diagnosis, we could expect subnotification of this condition on death certificates. For our estimate, we assumed a 46% rate of subnotification of SCD diagnosis in the death certificates, consistent with literature data; however, there is considerable uncertainty in this statistic.²⁸ Because this is not a birth cohort, and the overall SCD population is not directly measured, this is not a traditional survival analysis; our survival curves were modeled and not directly observed. Nevertheless, in the absence of national birth cohorts for this population, our estimates provide helpful information and, to the best of our knowledge, are probably the best estimates available. Furthermore, the model considered parameters from different sources, not allowing an adequate estimate of statistical uncertainty (ie, confidence or credible intervals). Finally, it is important to note that considering the data available we were not able to stratify SCD survival based on the SCD genotype, to access the impact of other risk factors for mortality, or to assess the impact of disease modifying therapies, such as hydroxyurea, on survival patterns.

In conclusion, our results highlight the burden of SCD in Brazil and the necessity for improved care for this population. Despite improvements in newborn screening programs for SCD, mortality remains high; in the short term, providing access to known effective treatments and adequate health care for complications should be a major priority.

Contribution: R.D.C., F.F.C., C.L., and A.C.S.-P. critically reviewed the manuscript; C.B.M. collected and analyzed data, and wrote the manuscript; M.F. conceived and designed the study, analyzed data, and critically reviewed the manuscript; H.F. and C.T.B. contributed to study design and critically reviewed the manuscript; and all authors approved the manuscript.

Conflict-of-interest disclosure: R.D.C., F.F.C., C.L., and A.C.S.-P. received honoraria from Novartis for manuscript reviewing. C.B.M. and M.F. received honoraria from Novartis through HTAnalyze for study design, data analysis, and manuscript writing. The support was funded by Novartis. H.F. and C.T.B. are employees of Novartis.

Correspondence: Homero C. R. Souza Filho, Department of Hematology–Oncology, Novartis Oncology, Avenida Professor Vicente Rao, 90 CEP 04605-000, São Paulo-SP, Brazil; e-mail: homero.filho@novartis.com.