Introduction
Over a decade after the development of a screening test for HIV, the reduction of the transmission of HIV by blood transfusion in sub-Saharan Africa remains a public health challenge. Screening of blood has been difficult in many African countries due to lack of resources for the purchase and distribution of equipment, reagents and the technical training required for an effective blood screening program [1,2]. In a survey of 733 hospitals in Zaire, only 29% received HIV testing kits on a regular basis and few centres had standard practices of the recording of HIV test results [3]. One year after implementation of blood donor screening, an estimated 25% of pediatric HIV infections were still caused by blood transfusions [4]. Even in areas that are able to screen blood, the rate of new infections in many African countries is high and HIV screening tests are unable to detect recently-infected blood donors who have not yet developed HIV antibodies [5,6].
Confronted with the continuing problems of blood safety in developing countries, the reduction of unnecessary transfusions has been emphasized as a practical and cost-effective method to reduce transfusion-associated HIV transmission [1,2,7-9]. This strategy is particularly important among children, who receive 50-67% of all transfusions given in many African countries [10-15]. In Tanzania, an audit of transfusion practices found that 75% of all avoidable transfusions were among children under 5 years of age [16]. Studies in Kenya and Côte d'Ivoire demonstrated that nearly half of the pediatric transfusions could have been prevented without increasing in-hospital mortality if transfusions had been limited to children with severe anemia (hemoglobin less than 5.0 g/dl in Kenya and less than 6.0 g/dl in Côte d'Ivoire) who demonstrated clinical evidence of cardiorespiratory decompensation [17,18]. However, these studies evaluated only in-hospital survival, and did not account for variations in the length of hospitalization, the effects of other underlying illnesses, and the risk of recurrent anemia and anemia-associated mortality after the period of hospitalization. We conducted this study in a malaria-endemic area of Kenya to evaluate the effect of transfusion on pediatric mortality and hematologic recovery both during and after hospitalization.
Materials and methods
Patients
The study was conducted as part of a series of studies on anemia, malaria, and child mortality at Siaya District Hospital (SDH), located in a rural area of western Kenya with intense, perennial transmission of Plasmodium falciparum malaria. Although 75% of P. falciparum infections in this area exhibit moderate to high levels of resistance (RII or RIII) to chloroquine therapy [19], chloroquine was often used as standard treatment for children with malaria.
All children who were admitted to the 50-bed pediatric ward from March to September 1991 had their hemoglobin measured by the study team at the time of admission. Each child with severe anemia (hemoglobin < 5.0 g/dl) was enrolled, along with the next child admitted with hemoglobin ≥ 5.0 g/dl (constituting a 33% sample of children with hemoglobin ≥ 5.0 g/dl).
At the time of enrollment, medical history and a venous blood sample were collected and a standardized physical examination was performed. Signs of cardiorespiratory compromise were recorded. For the purpose of analysis, 'respiratory distress' was defined as having intercostal or subcostal retractions, forced expiration (grunting), or nasal flaring. These were selected as clinical indicators for the need for increased oxygencarrying capacity that could be recognized in areas where diagnostic equipment may not be available. Young children frequently had tachypnea or tachycardia associated with fever or dehydration; these signs were not included as indicators of cardiopulmonary compromise due to their poor specificity. Diagnoses were assigned by the principal investigators according to standardized criteria.
Children were evaluated daily during hospitalization and a capillary blood sample was collected at discharge. Patients were requested to return for follow-up 4 weeks and 8 weeks after the date of their admission, when a medical history and capillary blood sample were collected. Patients were visited at their homes if they did not return for their scheduled follow-up. If children died between follow-up visits, the reported date of death was recorded. To standardize follow-up periods for the purpose of analysis, '4-week follow-up' included any visits between 15 and 45 days after the date of admission, and '8-week follow-up' applied to any visits more than 45 days after admission.
During hospitalization, all medical care, including the use of blood transfusion, was administered by hospital staff according to routine hospital practice. Whole blood, screened for HIV by the hospital blood bank, was used for transfusion. Hospital staff reported that they transfused children with hemoglobin < 5.0 g/dl. However, transfusions that were ordered were often delayed for days or not administered because blood or other supplies were not available.
Laboratory
Capillary hemoglobin was measured at the time of admission, discharge, and follow-up (HemoCue, Inc., Mission Viejo, California, USA). At enrollment, a venous blood sample was collected for complete blood count (Coulter Counter, Model M350; Coulter Electronics, Luton, Bedfordshire, UK), blood culture, and a thick and thin blood smear for malaria parasites. Malaria smears were stained and read by study team technicians. Serum was tested for HIV-1 by enzymelinked immunosorbent assay (Wellcozyme, Murex Diagnostics, Dartford, Kent, UK), and repeatedly reactive specimens were tested by Western blot. Cerebral spinal fluid was cultured from children with mental status changes (coma, lethargy, hypotonia or non-consolable irritability). Children with tachypnea, grunting, nasal flaring, intercostal retractions or wheezing received a chest roentgenogram.
Statistical analysis
Chi-square and Wilcoxon rank sum tests were used to evaluate variables associated with anemia and mortality among severely anemic children who were and were not transfused. To evaluate the effect of transfusion on mortality, we computed Kaplan-Meier estimates of the survival functions for severely anemic children, stratified by the presence of respiratory distress (grunting, flaring, or nasal retractions) and receipt of transfusion. Proportional hazards models were constructed to quantify the association between transfusion and survival, while controlling for hemoglobin, malaria parasitemia, respiratory distress, concomitant illnesses, and medical therapy. To evaluate the effect of transfusion on mortality according to underlying illness, interaction terms were assessed for each diagnosis with transfusion. Backwards elimination was used to eliminate variables that were neither confounders nor associated with mortality. The effectiveness of transfusion in reducing mortality was defined as 1 - HR, where HR is the hazard ratio for transfusion obtained from the proportional hazards model. The effectiveness can be interpreted as the percentage reduction in mortality obtained from transfusion.
To evaluate the effect of transfusion on hematologic recovery, we constructed a weighted longitudinal linear model of hemoglobin at each follow-up for all of the hospitalized children (PROC MIXED, SAS version 6.08, SAS, Cary, North Carolina, USA), according to receipt of transfusion, mortality, and malaria parasitemia on admission and follow-ups, while controlling for other factors associated with hemoglobin (e.g., age, underlying illness). The modelled hemoglobins for children who ultimately died were fitted with interaction terms, when significant, to allow for changes in hemoglobin that were different from children who survived throughout the follow-up period. This methodology generalized existing linear modelling techniques by allowing the user to specify the nature of the correlation between observations (in this case, patient follow-ups). Previous methodologies either forced an assumption of independence between patient visits or were quite limited in the types of correlational structures that could be assumed. Our models assumed an autoregressive (lag 1) correlation between successive patient visits. The final model and correlational structure were deemed acceptable, based on the close fit of the estimated least squares means to the observed cell means.
Results
Study population
A total of 1223 children were admitted to the pediatric ward during the study period. Of these, 303 (25%) had hemoglobin < 5.0 g/dl and were enrolled in the study, along with 303 with hemoglobin ≥ 5.0 g/dl. Ninety-six per cent of the children enrolled were under 5 years of age. All enrolled patients were monitored for the 8-week follow-up period except for 15 (5%) who had moved and could not be located after discharge.
Severe pediatric anemia
Children with hemoglobin < 5.0 g/dl were younger and more likely to have malaria parasitemia and clinical signs of respiratory compromise compared with children with hemoglobin ≥ 5.0 g/dl. Children with hemoglobin < 5.0 g/dl were admitted for various illnesses, but no admission diagnoses were significantly associated with severe anemia (Table 1).
Transfusion, malaria parasitemia, and hematologic recovery
Although hospital staff reported that they ordered transfusions for children with hemoglobin < 5.0 g/dl, 116 (38%) of these children were not transfused, most often due to shortages of blood and supplies. Characteristics of children with hemoglobin < 5.0 g/dl who were and were not transfused are summarized in Table 1. In addition, transfusions were often delayed; 34% of the transfused children received their blood 2 or more days after the transfusion was ordered.
Transfusion significantly increased the mean hemoglobin level. Among the 229 severely anemic children who were living at the time of discharge, the mean hemoglobin level at discharge of the transfused children (9.0 g/dl) was significantly higher than that of the children who were not transfused (5.8 g/dl, Wilcoxon rank sum test, P < 0.001). Although mean hemoglobin continued to rise after discharge in both the transfused and non-transfused groups, transfused children continued to have significantly higher mean hemoglobin levels at both the 4-week and 8-week follow-up visits compared with those who had not been transfused (Table 2).
Among the severely anemic children, hematologic recovery was significantly attenuated by the presence of malaria parasitemia at the time of follow-up, in both the transfused and non-transfused groups. For children without parasitemia on follow-up, the longitudinal models of all pediatric admissions demonstrated that transfused children continued to have significantly higher mean hemoglobin levels at both the 4-week and 8-week follow-up visits compared with those who had not been transfused (least-square means of hemoglobin at 4 weeks: 10.2 ± 0.21 versus 9.6 ± 0.15 g/dl, P = 0.01; at 8 weeks: 10.7 ± 0.27 versus 9.8 ± 0.18 g/dl, P < 0.01). However, the presence of malaria parasitemia at the time of follow-up negated the benefit of transfusion on hematologic recovery. Among children with parasitemia, on follow-up, there was no significant difference between the mean hemoglobin levels of those who were transfused and those not transfused at 4 weeks (8.8 ± 0.23 versus 8.9 ± 0.19 g/dl) and 8 weeks after admission (8.4 ± 0.22 versus 8.3 ± 0.18 g/dl). The same relationship between parasitemia and hematologic recovery was seen among children who survived or died during the study period (Table 3).
Transfusion and mortality
Of the 288 severely anemic children who completed the 8-week study, 30% (88) died: 16% (47) died during hospitalization, 10% (30) died between the time of discharge and the 4-week follow-up, and 4% (11) died between the 4- and 8-week follow-up visits. Children with severe anemia were significantly more likely to die during the study period than those with hemoglobin ≥ 5.0 g/dl (30% versus 19%, χ2 test, P < 0.01, Table 1).
Among severely anemic children, those who were transfused had a lower overall mortality rate (21%) than the children who were not transfused (41%, relative risk, 0.52; 95% confidence interval, 0.36-0.73, Table 2). Transfusion had the greatest impact on the survival of children with clinical signs of respiratory distress on admission (Fig. 1), and the benefit of transfusion in this population was limited to the first 2 days of hospitalization. When the analysis was restricted to those children with respiratory distress who survived beyond the first 2 days of hospitalization, there was no significant difference between mortality rates of transfused and non-transfused children during hospitalization or after discharge (Fig. 2, proportional hazards model, P = 0.55).
Severely anemic children without respiratory distress on admission had lower mortality rates than those admitted with respiratory distress, both during and after hospitalization (Fig. 1). In-hospital mortality rates among children without distress were low, regardless of whether they were transfused (none out of 52) or not transfused (three out of 56, Fisher exact test, P > 0.05). Among the 12 children without respiratory distress who died, nine died after hospitalization. Unlike those with respiratory distress, children admitted without respiratory distress who were transfused had a lower post-hospital mortality rate than those who were not transfused (3.8 versus 14.8%, respectively; χ2 test, P = 0.06). When the analysis was restricted to children without respiratory distress who survived beyond the first 2 days of hospitalization, transfusion was associated with decreased mortality (Fig. 2, proportional hazards model, P < 0.05). Among those without respiratory distress who survived beyond the first 2 days of hospitalization, patient characteristics, including underlying disease, age, parasitemia, and hemoglobin, were similar among patients who lived and died in the transfused and non-transfused groups.
Among all children with severe anemia, the mortality rates remained high in both the transfused and non-transfused groups after hospitalization. Evaluation of admission diagnoses revealed that severely anemic children often had serious concomitant illnesses: 63% had respiratory illness, 33% had malaria, 22% had malnutrition, 12% had bacteremia, and 8% were HIV-seropositive (Table 1). Regardless of underlying disease, the mortality rates were lower among children who were transfused compared with children who were not transfused. In the proportional hazards models, there were no significant interactions between the benefit of transfusion and diagnosis; transfusion did not confer differential benefit to children with different diseases, including malaria.
Among severely anemic children, the levels of hemoglobin at discharge was not predictive of risk of death after hospitalization. Although the mean hemoglobin level at discharge was significantly higher among transfused children, there was no significant difference between the mean hemoglobin level at discharge of children who did or did not survive after hospitalization in the transfused (9.0 g/dl, 9.5 g/dl, P = 0.56) and non-transfused groups (5.8 g/dl, 6.0 g/dl, P = 0.58).
Discussion
Defining the appropriate use of blood transfusion is an increasingly critical issue, particularly in developing countries, where severe anemia is widespread, the prevalence of HIV infection among blood donors is high, and blood screening is often inadequate. Little information, however, has been available to assist health care providers and policy makers to define the appropriate use of blood in Africa. This study systematically followed all severely anemic children admitted to a Kenyan hospital to evaluate their outcome during and after hospitalization and to determine the effect of transfusion on hematologic recovery and mortality.
Transfusion increased the mean hemoglobin level of severely anemic children during hospitalization and throughout the 8-week follow-up period in an area where malaria, nutritional deficiencies, and other infectious processes associated with persistent hematologic decompensation are prevalent. Malaria-associated anemia is mediated through immune-mediated hemolysis [20], dyserythropoiesis [21], and hypersplenism [22], resulting in shortened erythrocyte survival [23] that persists for up to 3 weeks after clearance of the malaria parasites [24,25]. Although we expected only a brief increase in hemoglobin after transfusion, the transfused children maintained a higher mean hemoglobin level throughout the 8-week study period than those who were not transfused. It is important to note that children who were not transfused also had a significant increase in hemoglobin during the follow-up period, suggesting that if appropriate interim treatment is provided, hemoglobin will increase in a short period of time without the use of blood.
Malaria parasitemia severely attenuated the hematologic recovery of both transfused and non-transfused children, as has been reported elsewhere among children with moderate anemia [19]. Chloroquine continues to be used for treatment of children with malaria in many areas of Africa where chloroquine resistant P. falciparum is prevalent. The expanded use of effective malaria therapy among anemic children is needed to decrease the alarming prevalence of anemia, anemia-associated mortality, and the use of blood transfusions.
Practitioners have often been reluctant to reduce the use of pediatric transfusions out of concern for hematologic decompensation after hospitalization. This longitudinal study supports previous cross-sectional studies conducted in Africa which demonstrated the benefit of transfusion, but only when given to severely anemic children with cardiorespiratory distress [11,12,17,18,26,27] and when given within the first 2 days of hospitalization [17]. Our results further emphasize the need for transfusion as a life-saving intervention to be given promptly to clinically decompensated children. This hospital, as well as many others in Africa [10], frequently does not have a supply of banked blood and relies on family members to provide donations. However, the time required to collect and screen blood from family donors typically does not provide blood to patients when it is needed to improve survival. Programs aimed at improving clinical, laboratory and blood bank services are needed to improve the use of transfusions and the immediate availability of blood.
Children without respiratory distress were at a lower risk of death than those with respiratory distress. No benefit of transfusion was found during hospitalization among children without respiratory distress, but lower mortality rates were observed among transfused children during the follow-up period. The hemoglobin levels at discharge, concomitant illness, malaria parasitemia, and parasite density were not predictive of mortality. In contrast, a randomized trial of transfusion among Tanzanian children with hemoglobins between 4.0 and 5.7 g/dl and who had no evidence of congestive heart failure or pneumonia, found no difference in 8-week mortality rates [28]. Additional studies are needed to clarify the indications for transfusion among children who do not demonstrate an acute need for blood; fluid replacement, malaria therapy, nutritional supplementation, and appropriate antibiotics may be more important interventions to promote long-term survival in this population.
Our study was conducted in the context of routine medical care in a rural Kenyan hospital. Transfusion was not randomly allocated, and there may have been bias in who did or did not receive transfusion. However, transfused and non-transfused groups had similar clinical characteristics, and many important potential confounders, such as hemoglobin level, malaria parasitemia, respiratory distress and underlying illness were controlled for in the analysis. Children received clinical and laboratory evaluations on admission and we do not know if their cause of death was the same as that identified at the time of admission. More information is needed to determine the causes of death among anemic children and how improved treatment and supportive care could further reduce the need for blood. Finally, this study was conducted in an area of intense, perennial transmission of P. falciparum where chronic and recurrent malaria infections were common. It is uncertain if our findings can be generalized to other areas where malaria transmission is more seasonal and hyperparasitemia may be more prevalent.
In the past decade, severe anemia has been increasingly recognized as an important contributor to mortality among young African children. This longitudinal investigation demonstrated that, regardless of transfusion, severely anemic children had higher mortality rates than children without severe anemia both during and after hospitalization; 15% of the transfused children and 17% of the non-transfused children died after hospitalization. In addition to promoting the safe and effective use of blood, programs are urgently needed that focus on the prevention, early diagnosis and effective treatment of the causes of anemia and concomitant illnesses in order to promote child survival and prevent the high rate of transfusion utilization.
Acknowledgments
We thank Drs D. Kleinbaum, D. Heymann, R. Yip and the Siaya District Hospital staff for their collaboration and assistance, the CDC/KEMRI study team for their hard work, and J. Ochien'g for effectively conducting the follow-up in unmapped, unmarked, remote areas. This paper is published with the permission of the Director, Kenya Medical Research Institute.
References
1. Heymann DL, Edström K: Strategies for AIDS prevention and control in sub-Saharan Africa. AIDS 1991, 5 (suppl 1):S197-S208.
2. Ryder RW: Difficulties associated with providing an HIV-free blood supply in tropical Africa. AIDS 1992, 6:1395-1397.
3. N'Ntita I, Mulanga K, Dulat C, et al.: Risk of transfusion-associated HIV transmission in Kinshasa, Zaire. AIDS 1991, 5:437-439.
4. Shaffer N, Hedberg K, Davachi F, et al.: Trends and risk factors for HIV-1 seropositivity among outpatient children, Kinshasa, Zaire. AIDS 1990, 4:1231-1236.
5. Belsey EM, Maskill W, Emmanuel J, et al.: Model of 'window' period for HIV-infected blood. IXth International Conference on AIDS, Berlin, June 1993 [abstract WS-C11-6].
6. Savarit D, De Cock KM, Schutz R, et al.: Risk of HIV infection from transfusion with blood negative for HIV antibody in a west African city. BMJ 1992, 305:498-502.
7. Addo-Yobo EOD, Lovel H: How well are hospitals preventing iatrogenic HIV? A study of the appropriateness of blood transfusions in three hospitals in the Ashanti Region, Ghana. Trop Doct 1991, 21:162-164.
8. Global Blood Safety Initiative: Guidelines for the appropriate use of blood. Geneva: World Health Organization, 1989, WHO/GPA/INF/89.18.
9. Foster S, Buvé A: Benefits of HIV screening of blood transfusions in Zambia. Lancet 1995, 346:225-27.
10. Jäger H, N'Galy B, Perriens J, et al.: Prevention of transfusion-associated HIV transmission in Kinshasa, Zaire: HIV screening is not enough. AIDS 1990, 4:571-574.
11. Gumodoka B, Vos J, Kigadye FC, van Asten H, Dolmans WMV, Borgdorff MV: Blood transfusion practices in Mwanza Region, Tanzania. AIDS 1993, 7:387-392.
12. Dorward JA, Knowles JK, Dorward IM: Treatment of severe anaemia in children in a rural hospital. Trop Doct 1989, 19:155-158.
13. Greenberg AE, Nguyen-Dinh P, Mann JM, et al.: The association between malaria, blood transfusions, and HIV seropositivity in a pediatric population in Kinshasa, Zaire. JAMA 1988, 259:545-549.
14. Lackritz EM, Ruebush II TK, Zucker JR, Adungosi JE, Were JBO, Campbell CC: Blood transfusion practices and blood-banking services in a Kenyan hospital. AIDS 1993, 7:995-999.
15. Addo-Yobo EOD, Love H, Aidoo M: Malnutrition and malaria gave him anaemia, then the blood transfusion gave him HIV [letter]. Trop Doct 1992, 22:123.
16. Vos J, Gumodoka B, Ng'weshemi JZ, et al.: Are some blood transfusions avoidable? A hospital record analysis in Mwanza Region, Tanzania. Trop Geogr Med 1993, 45:301-303.
17. Lackritz EM, Campbell CC, Ruebush TK, et al.: Effect of blood transfusion on survival among children in a Kenyan hospital. Lancet 1992, 340:524-528.
18. Lackritz EM, Djomand G, Vetter KM, et al.: Beyond blood screening: reducing unnecessary transfusions and improving laboratory services in Côte d'Ivoire. IXth International Conference on AIDS, Berlin, June 1993 [abstract WS-C11-2].
19. Bloland PB, Lackritz EM, Kazembe PN, Were JBO, Steketee R, Campbell CC: Beyond chloroquine: Implications of drug resistance for evaluating malaria therapy efficacy and treatment policy in Africa. J Infect Dis 1993, 167:932-937.
20. Greenwood BM, Stratton D, Williamson WA, Mohammed I: A study of the role of immunological factors in the pathogenesis of the anaemia of acute malaria. Trans R Soc Trop Med Hyg 1978, 72:378-385.
21. Abdalla S, Weatherall DJ, Wickramasinghe SN, et al.: The anaemia of P. falciparum malaria. Br J Haematol 1980, 46:171-183.
22. Looareesuwan S, Ho M, Wattanagoon Y, et al.: Dynamic alteration in splenic function during acute falciparum malaria. N Engl J Med 1987, 317:675-679.
23. Looareesuwan S, Davis TM, Pukrittayakamee S, et al.: Erythrocyte survival in severe falciparum malaria. Acta Trop 1991, 48:263-270.
24. Woodruff AW, Ansdell VE, Pettitt LE: Cause of anaemia in malaria. Lancet 1979, i:1055-1057.
25. Phillips RE, Looareesuwan S, Warrell DA, et al.: The importance of anaemia in cerebral and uncomplicated falciparum malaria: role of complications, dyserythropoiesis and iron sequestration. Q J Med 1986, 58:305-323.
26. Pape GL: Response to drug treatment in children with chronic anaemia. Trop Doct 1989, 19:46.
27. Republic of Uganda AIDS Control Programme, Ministry of Health:. Indications for blood transfusion in Uganda. Entebbe: Health Education Printing Unit, 1989.
28. Holzer BR, Egger M, Teuscher T, Koch S, Mboya DM, Smith GD: Childhood anemia in Africa: to transfuse or not transfuse? Acta Trop 1993, 55:47-51.
© Lippincott-Raven Publishers.