A ten year review of the sickle cell program in Muhimbili National Hospital, Tanzania

Tanzania has a population of 45 million and is classified as a low income country. The MSC program was established through collaboration between Muhimbili University of Health and Allied Sciences (MUHAS), the oldest and largest biomedical university in Tanzania and MNH, the national referral hospital which is in Dar-es-Salaam on the Eastern coast.

Individuals attending MNH who were known or suspected clinically to have SCD or related to SCD individuals were invited to the clinic for enrolment.

The hospital had two clinics a week for SCD patients. During the course of the program, a third clinic was started. Patients were encouraged to visit the clinic in the event of an acute illness. All individuals who were enrolled were given a unique identifier number and hospital case number. Case report forms (CRF) were completed including demographic, clinical history and physical examinations. Residence information, including telephone number and consent to contact the family in case the patients did not attend clinic for more than 9 months were obtained. Management of SCD followed hospital guidelines which included prescription of folic acid 5 mg/day, investigations to identify underlying additional causes of anaemia, blood transfusion (BT) for those with haemoglobin< 5 g/dL, or heart failure. Individuals with splenomegaly requiring BT (> 3 times a year) were referred for splenectomy. Oral iron was prescribed to those with iron deficiency, defined by low serum ferritin or empirically to those with low MCV (< 80 fl) and MCHC (< 25 g/dL). Use of insecticide treated nets was emphasized to prevent malaria infections and chloroquine prophylaxis was prescribed until its use was discontinued due to high resistance. From 2010, SCD children under 5 years of age were prescribed daily oral penicillin.

At enrolment visits, peripheral blood samples in EDTA were analyzed for full blood count (haemoglobin (Hb), red blood cell count (RBC), mean corpuscular volume (MCV), mean corpuscular haemoglobin (MCH), mean corpuscular haemoglobin Concentration (MCHC)], white blood cell (WBC) and platelet count (PLT) (Pentra 60, Horiba ABX, Kyoto, Japan; Sysmex XT2000i, Hyogo, Japan). Nucleated RBC could not be differentiated from neutrophils by the haematology analyzer. Reticulocyte count (New methylene blue method; Sysmex XT2000i, Hyogo, Japan) and fetal haemoglobin (HbF) levels by High performance Liquid Chromatography (HPLC) (BioRad, Hercules, CA, USA) were obtained. SCD diagnosis was made by HPLC and Haemoglobin electrophoresis (Helena, Sunderland, Tyne & Wear, UK). Bilirubin total (BIL-T), bilirubin direct (BIL-D)], lactate dehydrogenase (LDH), alkaline phosphatase (ALP) and creatinine were assayed using a chemistry analyzer (Roche Cobas Mira, New York, USA or Abbott Architect, New York, USA). Daytime peripheral oxygen saturation (SpO₂) was determined using a pulse oximeter (pulse oximetry; Nellcor, Pleasanton, CA, USA). Peripheral Oxygen desaturation was defined as < 95%.

Data were managed using MySQL database (Sun Microsystems Inc., Santa Clara, California, USA). Analysis was performed using STATA and R statistical software (http://​www.​R-project.​org/​). Continuous variables with normal distribution were compared using t-test or ANOVA test and skewed continuous variables were compared using Wilcoxon sign rank or Kruskal Wallis tests, while categorical data were compared using chi-square tests. In order to evaluate the trend of different laboratory parameters, the mean difference between non-SCD and SCD were compared across the age groups. Some mean differences were log transformed to improve visibility of parameters with high and low levels of the difference. Weighted logistic regression was used to determine association between binary response variable and explanatory variables (clinical and haematological factors) accounting for under representation of non-SCD individuals. A p-value < 0.05 was considered to be statistically significant.

The description included the age structure and pattern of enrollment. SCD diagnoses were classified into 2 categories: SCD was for SS and non-SCD for those with AS and AA. SCD were almost all β^S/β^S with a small fraction of S/β⁰ thalassaemia (estimated at 4% - unpublished data) and very rare β^S/HPFH. SC or CC disease was not encountered. The age of patients was calculated at date of visit from the date of birth and three age group categories were defined: < 5, 5–17 and ≥ 18 (adults) years. Description of clinical and laboratory features between SCD and non-SCD individuals was limited to individuals in age group 5–17 years for the following reasons; 1) < 5 years age group is subject to considerable physiological changes; 2) ≥18 years age group is a heterogeneous group and it is not clear whether it consists of individuals with mild disease who have survived the early mortality of severe disease but who are symptomatic and therefore attend MNH for healthcare. The ethnic group of an individual was from self-reported ethnicity which is patrilineal. If this information was missing, the ethnic group of the mother was used. We plotted the geographical distribution of SCD in Tanzania based on the ethnicity of the SCD cases registered at MNH. The region where a particular ethnic group resides was used as a surrogate for the geographical distribution of SCD. Geographical coordinates of Tanzania and regional administrative boundaries were obtained from Global administrative boundaries website [26].

Enrollment is a composite result of the variable prevalence of the Hb S gene, of migration patterns to Dar-es-Salaam and referral practices to MNH. With that proviso, we note that regions with the highest SCD cases were those along the coast (> 10 cases per 100,000 people), the North-Eastern part, which include regions of Coast, Tanga and Zanzibar and along the Lake Victoria zone (Tabora, Shinyanga, Kagera and Mara) with prevalence ranging from 5 to 12 cases per 100,000 people (Fig. 2). In contrast, Arusha, Manyara, Kilimanjaro and Rukwa regions had the lowest representation of SCD (< 2 cases per 100,000).

Table 2 compares the clinical and laboratory parameters between non-SCD and SCD individuals aged 5–17 years for the reasons outlined under Methods section. Compared to non-SCD, the SCD group had statistically significant higher pulse rate and mean body temperature, but lower levels of blood pressure and peripheral oxygen saturation. The prevalence of low oxygen saturation (< 95%), jaundice and splenomegaly was significantly higher in SCD than in non-SCD group. Regarding laboratory features, SCD is a well-defined disease for which there is no a priori appropriate control group. Comparison of SCD and non-SCD individuals enrolled in the MSC programme, we find that all parameters were significantly different. The SCD group had a significantly lower Hb (g/dl) (7.4 vs 10.6), alkaline phosphatase and creatinine but higher levels of reticulocyte count, RDW, HbF, MCV, MCH, WBC, platelet count, total, direct and indirect bilirubin, aspartate transaminase, and lactate dehydrogenase. Similar trends for parameter levels were observed even when analysis was extended for all age groups, Fig. 3. The trends of mean difference were similar across the age groups, except for MCHC (non-SCD aged < 5 years had higher levels) and ALP (non-SCD aged > 18 years had lower levels) when compared to other groups. Only mean difference for platelets, ALP and creatinine for children < 5 years were not significantly different between the non-SCD and SCD groups (p > 0.05).

Evaluation of the clinical and laboratory features in three different age groups within the SCD population is presented in Table 3 and Fig. 4. There were statistically significant differences between the three age groups and interestingly, many of the parameters showed monotonic trends with increasing age. Notably, children aged 0–4 years had a significantly higher pulse rate and prevalence of palpable spleen, whilst peripheral oxygen saturation was lower in older SCD patients and prevalence of jaundice increased with age. The blood pressure (both systolic and diastolic) was significantly higher in the older SCD populations. For 1165 patients (31%), we obtained medical history on the causes of hospitalization in the 12 month period preceding enrollment (Fig. 5). Overall, the three commonest causes of hospitalization were pain, fever (including malaria) and severe anemia. Severe anaemia was the leading cause in children < 18 years (36.9% in 0–4 and 30.7 in 5–17 years), whereas pain was the main cause amongst adults aged ≥18 years (43.2%). Pain (12.9%), fever (12.0%) and malaria (11.2%) contributed equally in children below 5 years, while in children aged 5–17 years pain (27.4%) was the second contributing factor. Out of 469 patients who were admitted with symptoms of severe anaemia as the primary cause, severe anaemia (hb < 5 g.dL) was confirmed in 13.3% of the patients, 36 (7.7%) had pain, while only one patient had jaundice. Contrarily, of patients with Hb results (338) among those admitted with pain as the primary cause, only 11 (3.2%) had confirmed severe anaemia, while only three (out of 388) had jaundice.

The levels of Hb, MCV, MCH and MCHC showed a slight but significant upward trend, whereas RDW decreased with age. Reticulocyte counts were elevated in all age groups but they decreased with age. There were significant differences in the WBC counts with counts decreasing with age whereas platelet counts increased with age. Total, direct and indirect bilirubin all increased significantly with age (p < 0.001). AST and ALP decreased with age, however the decrease was only significant for ALP (p < 0.001).

Most studies of SCD have focused on children and adults separately, reflecting the fact that most healthcare systems have dedicated but separate paediatric and adult services. The MSC program, instead, aimed to integrate the care of SCD patients across their life span. First, we noted that pain, infection and anaemia were the three most common causes of hospital visits (Fig. 5): but whereas in children the most common cause was anaemia, adults were more affected by painful episodes. This is similar to findings in other series, where 50% of adults experience painful episodes compared to only 10% in children [38]. Next, we analyzed by age groups all other data over the entire age span from infancy to adulthood. To our knowledge, this is one of the first studies in which such a systematic analysis has been attempted. Here we will focus only on a few observations which we think are important. First, mean level of Hb increased with age, though only slightly, and this was associated with a decrease in the mean reticulocyte count. This correlation is expected, as the increased reticulocyte count reflects a normal compensatory erythropoietic response to anaemia. The decreased reticulocyte count in older patients with SCD may also have reflected reduced erythropoietin levels in the context of the onset of chronic kidney disease, as evidenced by the higher creatinine in the older age group. An alternative explanation may possibly be a direct effect of vaso-occlusive events on the bone marrow resulting in reduced erythropoiesis with impaired reticulocyte production. Second, with respect to HbF, we expected that the mean HbF level would be highest in the youngest age group, since the physiological decrease in HbF from the level of about 80% at birth is much delayed in SS children – compared to controls – in at least the first 30 months of life [39]. However, the further decrease in HbF in the transition from the 5–17 age group to adulthood is somewhat paradoxical, because one would have expected that SCD patients with higher HbF would be, if anything, favored in terms of survival: since HbF was seen not to predict mortality in our previous work [33], this finding requires further research. Third, the mean values of red cell indices (MCV, MCH, MCHC) were lower in the younger age group which, in line with the fall in RDW with age, may mean that iron deficiency anemia is more common in the youngest patients (it is indeed common in Tanzania in the < 5 age group). Low MCV and MCH may also result from α-thalassaemia, which is generally regarded as an ameliorating factor in SCD; however, from these data and our previous work [33], low MCV does not seem to confer a survival advantage. Fourth, the mean WBC count decreased with age: which may suggest that the prevalence of infection is higher in children and decreases with age. However, the downward trend in WBC counts may also mean that the older age group is representative of less severe disease. In contrast, the mean platelet count increased with age, probably reflecting a decrease in splenic function: this explanation is supported by a lower prevalence of palpable spleen in the older age groups. The age-specific patterns in the level of WBC counts and platelets warrant further research, as the role of these blood components in the pathophysiology of SCD is being increasingly recognized [40]. Fifth, the mean level of bilirubin increases with age, in keeping with the observation that prevalence of jaundice increased with age. This may be explained by an increase in the rate of extravascular haemolysis; liver dysfunction is less likely since there was a decrease in the mean levels of ALP and AST (albeit not statistically significantly with respect to AST). We must also consider that the higher levels of ALP in the younger age groups could be related to increased bone tissue turnover associated with growth. A striking observation was the high level of LDH (more than 3 times the upper limit of normal) in all age groups, which suggests a significant level of intravascular haemolysis, although cardiac and liver dysfunction may also contribute. Further research is required to quantitate the age-specific rate of haemolysis and whether there is a change with age in the relative rates of intravascular versus extravascular haemolysis. Finally, the mean creatinine level increased with age, although most values remained within the normal range. The creatinine level on its own is not a sensitive index of early renal disease, but one wonders whether in individual cases the trend of creatinine with age may be a prelude to renal insufficiency, a known complication of SCD. It should be noted that the age-related trends in clinical and laboratory parameters in this report were also reported by Aliyu et al. [41] in Nigeria (208 individuals with SCD; Youngest 10 years; 7% > 35 years).

In one of the first systematic studies of SCD patients, Nagel and colleagues [42] noted that the β^S haplotype may account for the great variability in the clinical symptomatology and severity of SCD. In this population in Tanzania, this layer of heterogeneity is largely removed, since nearly all patients are homozygous for the Central African Republic (CAR) haplotype (previously referred to as Bantu haplotype) [43]: However, the spread of values remains just as great (Fig. 4); for example, the mean hematocrit level that Nagel et al. [42] reported for SS patients with CAR haplotype was 23.1, which corresponds almost to the mean Hb of 7.4 in the Tanzania population with the same haplotype, although the range in Tanzania is wider. Although the β-globin haplotype is regarded as a major determinant of HbF production, we found (Fig. 4) a markedly skewed distribution, whereby the top 2.5% of patients had HbF values some 4 times higher than the mean. Genomic studies conducted in SCD in Tanzania have already reported variability in prevalence of and co-existence of genomic loci that influence HbF, α-thalassemia, and G6PD levels [25, 37]. Further research is required, and is ongoing, to increase understanding in the complex interplay of genetic, environmental and socio-economic factors influencing variability of disease spectrum [44].

Because enrolment into the SCD cohort was hospital-based, the study may not have captured individuals with mild disease. The non-SCD group was not a population-based age- and sex-matched control group: rather, it was a ‘convenient’ comparison group consisting of patient family members and of individuals enrolled among those suspected to have SCD: hence they may have other medical conditions. Indeed, the clinical and laboratory features were often not normal. We did not separate controls with HbAA from those with HbAS. This is broadly justified because sickle cell trait subjects have no cells containing exclusively HbS and are haematologically normal. However it is possible that some parameters may differ in AA versus AS subjects: this requires further study. Finally, the work reported here is that from cross-sectional analysis of the enrolment cohort. Although the study was conducted over a 10 year period, we are not yet presenting longitudinal data.