Post-treatment haemolysis in African children with hyperparasitaemic falciparum malaria; a randomized comparison of artesunate and quinine

Parenteral treatment with artesunate has been shown to reduce mortality from severe malaria by 35% in adults and 23% in children compared to quinine [1, 2] and is the recommended first-line treatment for severe malaria [3]. Artesunate is well tolerated, safe and has fewer adverse effects than quinine. Until recently the only serious adverse event observed has been a Type-1 hypersensitivity reaction in approximately one per 3000 treatments [4‐6]. However, there are now several reports, mainly from temperate countries, of late haemolytic anaemia occurring 1 to 3 weeks after parenteral artesunate treatment of hyperparasitaemic falciparum malaria in returned travellers. This has raised general concerns about the safety of these life-saving compounds. Most cases followed intravenous artesunate, although some were reported after intramuscular artemether, intrarectal artesunate and oral artemisinin derivatives [7, 8]. There were no deaths, but blood transfusion was often required. A recent review of published data, from non-immune and semi-immune patients, estimated the incidence of late haemolysis after intravenous artesunate to be 13% (95% CI 9–18%) and the requirement for a blood transfusion at 9% (95% CI 6–14%) [7], although most data are from case reports and case definitions have varied substantially. But returned travellers constitute only a very small fraction of the global burden of severe malaria. Are these observations in travellers of clinical relevance to African children, who carry more than 90% of the global malaria disease burden, and for whom life-threatening anaemia is a common clinical presentation of falciparum malaria? Severe malarial anaemia in childhood is associated with significant mortality both in hospital and after discharge [9]. As haemolytic anaemia is the pathological hallmark of all malaria infections, and severe anaemia is a common feature of severe malaria, dissecting disease from drug-associated mechanisms is difficult. To investigate the pathogenesis of anaemia in African children and whether post-artesunate haemolytic anaemia is a significant problem, a randomised comparison was conducted of the immediate and delayed haematological responses to parenteral artesunate and quinine in children hospitalised with high falciparum parasitaemias.

The trial was conducted at the Kinshasa Mahidol Oxford Research Unit, Kinshasa, the Democratic Republic of Congo. Malaria in this area is endemic and perennial, with minor seasonal variations.

Children with acute falciparum malaria were eligible for enrolment if they were aged between 6 months and 14 years old, had body weight > 5 kg, haemoglobin ≥5.0 g/dL and parasitaemia ≥100,000 falciparum parasites/μL. Patients with any sign of severe malaria [3] (although hyperparasitaemia alone was not an exclusion criterion [10]), a history of hypersensitivity or contraindication to the use of artesunate or quinine, a clear history of adequate antimalarial treatment within 24 h, or the presence of a significant other illness, were excluded. The study was explained to caregivers in French or Lingala and they were requested to sign a consent form to allow their child’s participation in the study.

The study was an open-labelled randomised comparison of the hematological effects of parenteral antimalarial treatment with either artesunate or quinine. Computer randomisation, in blocks of 20, was prepared by a study statistician. Treatment allocation was concealed in sequential opaque envelopes prepared by an independent person. The intervention was assigned by the study nurse, after the doctor confirmed eligibility and the caregivers had signed the informed consent. After a full clinical examination, patients were randomised to receive intravenous (IV) artesunate (AS) 2.4 mg/kg body weight bolus (Guilin Pharmaceutical, Guangxi, China) immediately and at 12, 24, 48 and 72 h or quinine dihydrochloride (QN) 20 mg salt/kg body weight (Sterop, Brussels, Belgium) infused over 4 h, followed by 10 mg/kg body weight every 8 h infused over 2 h until 72 h.

Parenteral treatment was followed by a full course of oral artemether-lumefantrine (Coartem®, Novartis); the first dose was directly observed and the remaining doses were given with instructions to the caregiver for administration at home. Patients were followed-up weekly for 6 weeks. Any recurrent episodes of malaria were treated with fixed-dose amodiaquine-artesunate tablets (Winthrop®, Sanofi). Blood transfusion was given when haemoglobin was ≤5 g/dL regardless of clinical symptoms, or if the haemoglobin was ≤7 g/dL and there were one or more clinical features of severity present [3].

Malaria blood films were prepared at admission, 0 (baseline), 6, 12 h, then every 12 h until 2 consecutive microscopy negative blood films and at each follow-up visit. Parasites were counted using standard methods. Hemoglobin (Hb) and haematocrit (Hct) were assessed at the same time-points as the blood films using HemoCue Hb201+ (Angelholm, Sweden) and Hawksley Haematospin 1400 (Hawksley & Sons, Ltd. UK). Reticulocytes were counted as a percentage of total RBCs at 0, 72 h and weekly.

Ring-infected Erythrocyte Surface Antigen (RESA)-positive, falciparum-negative RBCs (once infected or oi-RBCs) were counted per 1000 RBCs at 0, 24, 48, 72 h and then weekly by immunofluorescence microscopy [11]. Total and differential white blood cell (WBC) counts were measured at 0, 72 h and weekly by QBC Star™.

Hemoglobin HbS [rs334] and A⁻ glucose-6-phosphate dehydrogenase deficiency defined by G6PD 376 (rs1050829) and G6PD 202 (rs1050828) were characterised from genomic DNA at the Wellcome Trust Centre for Human Genetics using Agena MassArray IPLEX platform and multiplexed PCR assay [12].

The prespecified primary endpoint was the proportion of patients who developed late anaemia, defined as a reduction of ≥10% in haemoglobin compared with any previous measurement between days 7 and 42 after starting parenteral antimalarial treatment. Secondary endpoints included: the proportions of patients requiring blood transfusion, the fractional decreases in haemoglobin in the first week of treatment, plasma LDH as a measure of haemolysis, reticulocyte and pitted erythrocyte counts, parasite clearance times, and frequency of serious adverse events.

The proportion of intravenous artesunate treated patients with delayed anaemia was assumed to be 20% [13]. To provide 90% power and 95% confidence to detect a difference of 20% vs. 4%, with 10% loss to follow-up, 108 children per arm were recruited.

Comparisons between groups used the Student’s t-test or Mann-Whitney U test, depending on distributions. Kendall’s Tau (τ) or Pearson’s correlation coefficient (rho_p) assessed associations between continuous variables. For more than two groups, the Kruskal-Wallis test was used. AUCs were estimated using the trapezoid rule. Hemoglobin, oi-RBC and parasite count values were censored on the day of malaria recrudescence or blood transfusion.

Parasite clearance was assessed as the time for the parasite count to fall by 50% (PC₅₀), the parasite clearance half-life (PC_1/2) [14] and the interval until patients became parasite negative by microscopy (PCT).

Clearance of oi-RBC was estimated by modelling the decline in log-transformed oi-RBC over time in a random-effects model, with variable intercept and slope. Mono- and biexponential decay curves were considered. The final model was selected based on the Akaike Information Criterion (AIC). Levels of oi-RBC below the detection limit were excluded from the analysis.

Safety reporting was performed according to the ICH Harmonized Tripartite Guideline for GCP (1996). The study was approved by the Oxford Tropical Medicine Ethics Committee and the University of Kinshasa, School of Public Health institutional review board. Data were entered using MACRO InferMed and analysed using STATA v14.0. (ClinicalTrials.​gov; Identifier: NCT02092766).

Between June 2014 and March 2015, 217 children were enrolled (Fig. 1). One study patient, whose parents withdrew their consent the first day of hospitalization, was replaced (AS). Four patients developed severe malaria in the first 24 h (AS 1; QN 3); for those allocated to QN, the treatment was changed to IV AS. In three other cases (QN), difficulties in obtaining venous access prompted a change to intramuscular (IM) artesunate. Thus, hematological outcomes were analysed in 211 children (AS 109; QN 102). The two groups had similar demographic, clinical and laboratory characteristics on admission (Table 1). Five patients were lost to follow-up after day 7 (AS 2; QN 3), all remaining patients were followed until day 42. Most patients (75.6%) had an adequate clinical and parasitological response at day 42; forty-one (18.9%) had a recurrent episode of malaria with a similar rate in the two arms (log-rank test p = 0.80) (Additional file 1: Table S1).

Anaemia is a consistent feature of falciparum malaria and results from the obligatory lysis of infected erythrocytes at schizogony, increased splenic clearance of both infected and uninfected erythrocytes and dyserytropoiesis. The relative contributions of these factors vary with age, immunity and disease severity. In tropical areas malaria anaemia may be compounded by haematinic deficiencies, hypersplenism, and genetic red cell disorders. As severe malarial anaemia can be fatal, it is important to identify preventable contributors. In this study of African children with acute falciparum malaria and high parasite densities living in a high transmission area post-artesunate haemolytic anaemia was substantially less common than reported in non-immune returned travellers.

Post-artesunate haemolytic anaemia has been attributed to the increased pitting of drug-damaged malaria parasites from infected erythrocytes [17]. Pitting is the physiological process whereby the spleen removes intraerythrocytic particulate matter, in this case the pyknotic ring-stage intra-erythrocytic parasites, and then returns the resealed red blood cells into the systemic circulation [11]. The life-saving benefit of artesunate over quinine derives from its rapid parasiticidal activity against the circulating ring-stage parasites. This results in a higher rate of parasite clearance and a correspondingly greater production of oi-RBC (Fig. 2b). These once-infected red blood cells (oi-RBC) have a much shorter survival time (7–14 days) compared with normal erythrocytes in healthy subjects (120 days) or in patients following severe malaria (44 days) [18, 19]. The shorter survival of oi-RBC following artesunate probably reflects drug killing and then pitting of developed ring forms whereas background (and quinine associated) pitting may only occur for very young ring stages shortly after merozoite invasion with correspondingly less damage to the erythrocytes. The rapid and synchronous elimination of these oi-RBC from the circulation 1 to 2 weeks after the start of antimalarial treatment might result in haemolytic anaemia [17].

In this study, a similar proportion of artesunate and quinine treated patients had a ≥ 10% reduction in haemoglobin, but the degree of anaemia was moderate, although there are small but significant differences in the time course. Patients receiving artesunate had slightly lower initial reductions in haemoglobin and delayed reticulocyte responses. Previous studies also indicated a clinically insignificant temporary suppression of reticulocytosis by artemisinins in the treatment of malaria [20, 21]. More than three times as many ring stage parasites were pitted after treatment with artesunate compared to quinine, with over half returning to the circulation, largely explaining the initially smaller reduction in haemoglobin in artesunate recipients. The ratio of oi-RBC to the initial parasitaemia (59%) was comparable to that observed in Malian children (60%) following artesunate, but lower than in returned travellers (78%) [17]. Moreover, in artesunate-treated travelers around 60% of originally parasitized cells were still present in the circulation 7 days after treatment, while in this study only approximately one quarter of the oi-RBCs were left at this stage, thereby reducing the number of erythrocytes contributing to late haemolysis. These findings suggest that other, presumably immune mediated, mechanisms leading to infected cell removal contribute to splenic clearance of parasitaemia in African children. The erythropoietic response to malarial anaemia was rapid (peak 7 days) [22], and took place earlier than in patients with less background immunity (14–21 days [17, 19, 23]). Hence in these semi-immune African children the lower contribution of pitting to parasite clearance after artesunate and the earlier reticulocyte response mitigated the loss of haemoglobin associated with the late lysis of oi-RBC, whilst more rapid clearance of pitted cells reduced haemolysis occurring after 7 days. Only one artesunate treated patient had haemolytic anaemia severe enough to warrant blood transfusion. This patient had a complicated disease course with suspected concomitant bacterial sepsis treated. Whether the haemolysis in this case was explained only by malaria and artesunate, or whether the presumed sepsis contributed cannot be determined.

Severe delayed haemolytic anaemia was observed in less than 1% of Congolese children with acute falciparum malaria and high parasitaemias, which is similar to the frequency reported in West African children with severe malaria [22]. A limitation of this study is that the patients were not clinically severe whereas post-artesunate delayed haemolysis has been reported particularly after severe malaria. Children with severe malaria could not be randomised because artesunate is clearly the better treatment. Nevertheless, the patients studied were at the severe end of the clinical spectrum of hospitalised uncomplicated malaria. Plasma PfHRP2 concentrations on admission, an indicator of parasite burden, were between the values reported for African children with uncomplicated and severe malaria (approximately 150 ng/ml and 1500 ng/ml respectively [24]).