Clinical Predictors of Severe Malarial Anaemia in a Holoendemic Plasmodium falciparum Transmission Area

SUMMARY

Severe malarial anaemia (SMA) is a common complication of Plasmodium falciparum infections, resulting in mortality rates that may exceed 30% in paediatric populations residing in holoendemic transmission areas. One strategy for reducing the morbidity and mortality associated with SMA is to identify clinical predictors that can be readily recognized by caregivers for prompt therapeutic interventions. To determine clinical predictors of SMA, Kenyan children (3-36 mos., n=671) presenting with acute illness at a rural hospital in Siaya District were recruited. Demographic, clinical, laboratory and haematological parameters were measured upon enrolment. Since HIV-1 and bacteraemia promote reduced haemoglobin (Hb) concentrations, children with these infections were excluded from the analyses. Children with P. falciparum mono-infections (n=355) were stratified into three groups: uncomplicated malaria (Hb≥11.0 g/dL); non-SMA (6.0≤Hb<10.9), and SMA (Hb<6.0 g/dL). SMA was characterized by a younger age, monocytosis, thrombocytopaenia, reticulocytosis, reduced erythropoiesis, elevated pigment-containing monocytes (PCM), respiratory distress, conjunctival and palmar pallor, splenomegaly, signs of malnutrition, and protracted fever and emesis. Logistic regression analysis demonstrated that age, reticulocyte count, presence of PCM and conjunctival and palmar pallor were significant predictors of SMA. Recognition of these clinical signs in children residing in resource-poor settings may help guide the identification and management of SMA.

INTRODUCTION

Greater than 80% of malaria-related mortality occurs in sub-Saharan Africa due to infections with Plasmodium falciparum (Berkley, et al 2005b, Snow, et al 1999b, WHO 2005). The majority of P. falciparum-related mortality occurs in immune-naïve infants and young children, accounting for 18% of all deaths before five years of age (Snow, et al 1999b, WHO 2005). Clinical manifestations of severe falciparum malaria vary according to transmission intensity and typically present as one or more life-threatening complications, including: hyperparasitaemia; hypoglycaemia; cerebral malaria (CM); severe malarial anaemia (SMA) and respiratory distress (Dzeing-Ella, et al 2005, English, et al 1998, Mockenhaupt, et al 2004). In holoendemic transmission areas, SMA is the primary clinical manifestation of severe childhood malaria, with CM occurring only rarely (Bloland, et al 1999, Ong’echa, et al 2006).

SMA is responsible for ~30% of the mortality in children less than three years of age in holoendemic P. falciparum transmission areas (Obonyo, et al 2007a) and, therefore, requires prompt recognition and treatment. In resource-poor settings, laboratory investigations are often limited, with diagnosis relying on medical history and physical examination (Chalco, et al 2005). However, the risk factors and clinical predictors of SMA have not been fully elucidated.

In sub-Saharan Africa, SMA may develop in children already debilitated by other causes of paediatric anaemia, such as nutritional deficiencies, haemoglobinopathies, bacteraemia, and human immunodeficiency virus-1 (HIV-1) (Berkley, et al 1999, Ong’echa, et al 2006, Otieno, et al 2006, Calis, et al 2008). These co-morbidities are common in African children and likely impact on the development and outcomes of malarial anaemia by altering the host immune response. We have recently shown that HIV-1 positive children have significantly more SMA during acute P. falciparum infections (Otieno, et al 2006). Moreover, bacterial pathogens, such as non-typhoid Salmonella, are common causes of bacteraemia in febrile, hospitalized children in sub-Saharan Africa and account for increased anaemia severity (Berkley, et al 1999, Graham, et al 2000, Walsh, et al 2000).

Since a number of previous studies contained an unknown prevalence of HIV-1 and bacteraemia (Luckner, et al 1998, Kahigwa, et al 2002, Verhoef, et al 2002, Ong’echa, et al 2006), it is difficult to interpret the impact of clinical, nutritional, demographic and socioeconomic factors on malarial anaemia severity. Although it is apparent that HIV-1 and bacteraemia can contribute to more profound anaemia in children with malaria, it is important to define SMA as a single disease entity so that gaps in our current knowledge about the patho-physiological basis of SMA can be advanced. As such, we determined the demographic, clinical and haematological parameters associated with SMA using a hospital-based cross-sectional study design in a holoendemic P. falciparum transmission area of western Kenya. Children with P. falciparum parasitaemia and varying severities of anaemia (3-36 mos. of age) were recruited at their first hospital contact for the treatment of febrile illness and all children found to have HIV-1 or bacteraemia, and those with CM or non-falciparum malaria, were excluded from the analyses.

PATIENTS, METHODS, AND MATERIALS

Study design and participants

Children were recruited from the paediatric ward of the Siaya District Hospital (SDH) in Siaya District, Nyanza Province, western Kenya, a holoendemic P. falciparum transmission area with an entomologic inoculation rate of 100-300 per annum (Beier, et al 1994). The prevalence of anaemia in Siaya District ranges between 60-90% in children less than five years of age (Bloland, et al 1999, McElroy, et al 2000). A detailed description of the study area, clinical facilities and study participant population is presented in our previous publication (Ong’echa, et al 2006).

Children of both genders (n=355, males=194 and females=161, aged 3-36 mos.) with fever and a positive blood film for P. falciparum (upon screening) were enrolled in the study from March, 2004 to January 2006. Children with a history of prior hospitalization (for any reason), those that received anti-malarial treatment within the previous two weeks, children with CM (an infrequent occurrence in western Kenya), and those with malaria from non-falciparum species were not eligible for participation. Since HIV-1 and bacteraemia are common co-pathogens that influence anaemia status in children with malaria (Berkley, et al 1999, Ong’echa, et al 2006, Otieno, et al 2006, Calis, et al 2008), all study participants were screened for HIV-1 and bacteraemia according to our previous methods (Ong’echa, et al 2006, Otieno, et al 2006, Ouma, et al 2008), and those found to be positive were excluded from the analyses. Pre- and post-test HIV counselling was provided to the parents/guardians of all participating children.

Based on haemoglobin (Hb) measurements, children with any density P. falciparum parasitaemia were placed into the following categories: uncomplicated malaria (Hb≥11.0 g/dL); non-SMA (6.0 g/dL≤Hb<11.0 g/dL); and severe malarial anaemia (SMA, Hb<6.0 g/dL). Since precise definitions of childhood anaemia differ according to geographic regions, information from a previous study that performed >14,000 longitudinal Hb measurements in children (< 48 mos.) in western Kenya (McElroy, et al 1999) was used to appropriately define SMA as Hb<6.0 g/dL. To place the study into a broader context, data are also presented according to the World Health Organization (WHO) definition of SMA (Hb<5.0 g/dL with any density parasitaemia) (WHO 2000).

Treatment was prescribed according to Ministry of Health, Kenya (MOH) guidelines, which included Coartem® (an Artemether and Lumefantrin combination) for uncomplicated malaria, intravenous quinine for severe malaria, and supportive care (i.e., haematinics and blood transfusions). Human subject approval was obtained from the University of Pittsburgh, USA; University of New Mexico, USA; and KEMRI, Kenya.

Clinical assessment

History of present illness was obtained from the child’s parent/guardian to assess findings indicative of malaria severity (Warrell, et al 1990, Marsh, et al 1995). Fever was defined as axillary temperature >37.5°C, and duration of fever was recorded as none, 1-3 and 4-14 days. Hypotension, tachycardia and tachypnea were adjusted according to age (Marx 2002). Respiratory distress was defined as the presence of any of the following signs: alar flaring, chest retraction, use of accessory muscles during respiration, or abnormally deep, acidotic breathing (Marsh, et al 1995). Dehydration was defined as presence of sunken eyes or fontanelles, dry mucosa or decreased skin turgor. Nutritional status was determined by assessment of visible malnutrition, coded as “poor or fair” vs. “good” (Berkley, et al 2005a), while height-for-age (stunting), weight-for-height (wasting), weight-for-age (underweight), mid upper-arm circumference (MUAC)-for-height, MUAC-for-age and head circumference-for-age Z-scores were determined according to our previous methods (Ong’echa, et al 2006). Markers of nutrition were therefore clinical rather than laboratory-based.

Laboratory investigations

Venous blood samples (<3.0 mL) were collected in EDTA-containing tubes at the time of enrolment, prior to initiation of supportive care or other treatment interventions. Presence and density of malaria parasites, and haematological measures, including full blood count, reticulocyte count and reticulocyte production index (RPI) were performed according to our previous studies (Were, et al 2006). The presence, number and density of malarial pigment (haemozoin)-containing monocytes (PCM) and neutrophils (PCN) were also determined per previous methods (Nguyen, et al 1995, Lyke, et al 2003, Awandare, et al 2007). To ensure accurate laboratory results, 8% of the samples were selected based on a random number generating algorithm and re-evaluated for all preceding measures. Peripheral blood smears (n=81) were also randomly selected (using a random number generator) from all groups for examination of RBC and WBC morphology, since some haematological characteristics are not detected with the automated methods. RBCs were examined for presence of sickle cells, leptocytes, schistocytes and spherocytes, while WBCs were assessed for lymphocytes and toxic granulations. Glucose levels were obtained with an Accu-check Compact® (Roche Diagnostics, Indianapolis, IN) glucometer. Haemoglobin variants and glucose-6-phosphate dehydrogenase (G6PD) deficiency were determined as described previously (Ouma, et al 2009). ABO phenotypes were determined by agglutination of RBCs with anti-A, anti-B, and anti-AB antisera (forward grouping).

Statistical analyses

Data were analyzed using SPSS (version 15.0, Chicago, Illinois, USA). Pearson’s rank χ² test was used for comparing proportions. Group medians were compared using Kruskal–Wallis test followed by pairwise Mann–Whitney U test when significant Kruskal–Wallis tests were obtained. Logistic regression was used to examine the clinical predictors of SMA (statistical significance set at P≤0.05). All variables significantly associated with anaemia severity in the descriptive analyses, and in the comparisons among proportions, were divided into three clusters (laboratory parameters, patient history and physical examination). For each of these three separate clusters, a binomial logistic regression analysis was conducted. The purpose of these cluster-specific regressions was to reduce the overall number of predictors in the final predictor set and thereby reduce the probability of committing Type I errors. All variables from the cluster-specific regressions that were significant at P≤0.05 were then examined, along with age, as predictors of SMA in two additional second-step regression analyses. Non-SMA (alternatively coded as Hb=6.0-10.9 or 5.0-10.9 g/dL) and SMA (alternatively coded as Hb<6.0 or <5.0 g/dL) were used as categorical dependent variables. For both regression models, non-SMA was used as the comparison group since the uncomplicated malaria group was excluded from the regression analyses due to small sample sizes.

RESULTS

Demographic and laboratory parameters

Of the 671 subjects presenting at hospital with acute febrile illness and eligible for enrolment, children with bacteraemia (n=191), HIV-1 (n=24) and absence of P. falciparum parasitaemia (n=101) were excluded from the analyses. The remaining P. falciparum(+) children (3 to 36 mos., n=355) were stratified into three clinical categories: uncomplicated malaria (UM, Hb≥11.0 g/dL, n=23); non-SMA (Hb 6.0-10.9 g/dL, n=200); and SMA (Hb<6.0 g/dL, n=132). Age and parasitaemia differed significantly across the groups, while none of the other parameters reached significance (Table I).

Table I.

Demographic, laboratory, haematological and genetic characteristics

	UM	Non-SMA	SMA	P	n
Demographic and Laboratory Indices
Age (mos.)	9.0 (9.0)	10.5 (11.0)	8.0 (7.0)	0.002	355
Female gender, no. (%)	11 (48)	86 (43)	64 (48)	0.599	355
Peripheral parasitaemia (/μL)	55,800 (115,055)	26,956 (55,984)	32,384 (66,997)	0.026	355
Geomean parasitaemia (/μL)	48,675	21,595	22,709		355
HDPa no. (%)	20 (87)	132 (66)	83 (63)	0.079	355
Hypoglycaemiab no. (%)	2 (13)	21 (20)	19 (17)	0.768	230
Erythrocytic Indices
Haemoglobin (g/dL)	11.5 (0.8)	7.8 (2.2)	4.9 (1.1)	<0.001	354
Haematocrit (%)	34.3 (3.0)	24.2 (5.5)	15.9 (4.2)	<0.001	354
RBC (x106/ml)	4.8 (0.9)	3.6 (1.1)	2.2 (0.8)	<0.001	354
MCV (fL)	75.6 (9.5)	68.8 (10.2)	70.4 (13.3)	0.009	354
MCH (pg)	24.8 (3.3)	22.2 (3.4)	22.3 (4.0)	<0.001	354
MCHC (g/dL)	33.4 (1.8)	32.2 (2.2)	31.9 (2.7)	<0.001	354
RDW (%)	17.0 (4.5)	20.7 (4.5)	23.1 (5.3)	<0.001	354
Reticulocyte count (%)	1.1 (1.8)	2.6 (3.6)	3.7 (5.9)	<0.001	354
RPI <2c	NA	144 (72)	105 (81)	0.052	329
Leukocytic Indices
WBC (×109/L)	13.5 (8.1)	10.3 (4.8)	13.0 (7.7)	0.001	353
Granulocytes (×109/L)	8.1 (6.5)	4.3 (3.3)	4.4 (4.0)	0.001	353
Lymphocytes (×109/L)	4.2 (2.1)	5.3 (3.2)	6.5 (5.1)	<0.001	353
Monocytes (×109/L)	0.7 (0.8)	0.9 (0.6)	1.3 (1.2)	<0.001	353
Platelet Indices
Platelets (×109/L)	247 (168)	159 (133)	144 (81)	0.001	354
MPV (fL)	7.4 (1.0)	8.3 (1.7)	8.6 (1.7)	<0.001	351
Genetic Characteristics
Sickle-cell traitd, no. (%)	3 (13)	37 (18)	14 (11)	0.140	355
G6PD deficiency, no. (%)	0 (0)	11 (6)	9 (7)	0.422	333
Blood group O, no. (%)	12 (52)	87 (43)	52 (40)	0.502	354
Haemozoin-containing Leucocytes
PCM, no. (%)	2 (9)	65 (33)	92 (70)	<0.001	355
PCN, no. (%)	2 (9)	11 (6)	13 (10)	0.320	355
PCMe (/μL)	0.4 (0.2)	0.4 (1.6)	0.5 (3.1)	0.072	159
PCN (/μL)	0.1 (0.1)	0.1 (0.9)	0.2 (0.4)	0.489	26

Data are presented as the median (interquartile range) unless otherwise noted. Differences between the groups were compared using Kruskal-Wallis test for medians, while proportions were compared using Pearson’s χ² test.

^aHigh density parasitaemia defined as ≥10,000 parasites/μL

^bBlood glucose <4.2 mmol per liter (75 mg/dL).

^cThe difference between the non-SMA and SMA groups (only) was compared, since an RPI<2 is not valid in the non-anaemic uncomplicated malaria group.

^dPresence of haemoglobin AS by electrophoresis. Difference between SMA and Non-SMA groups was significant (P = 0.025, Mann-Whitney U)

Haematological characteristics

Haematological characteristics of the study participants are shown in Table 1. Based on a priori grouping according to anaemia status, Hb, haematocrit and RBCs differed across the groups. MCV, MCH and MCHC were highest in UM, while RDW was highest in SMA. Although anaemic individuals (non-SMA and SMA) displayed reticulocytosis, the number of children with an RPI<2 [deficient erythroid proliferation/maturation (Hillman and Finch 1996)] was elevated in SMA (81%) vs. non-SMA (72%, P=0.052). Mild leukocytosis was present in the UM and SMA categories. Children with UM had ~two-fold higher neutrophil counts, while lymphocyte and monocyte counts increased with anaemia severity. Eosinophilia was rare (data not presented) and basophils were not recorded. Platelet counts progressively declined, while MPV increased with anaemia severity. Examination of randomly selected peripheral blood smears (n=81) revealed the following: leptocytes, schistocytes, acanthocytes and spherocytes were similar across the groups; macrocytosis was rare; and sickled cells were not detected. Nucleated RBC were observed in 4 of 17 (23.5%) SMA cases compared to 2 of 44 (4.5%) in the non-SMA group (P=0.053). Reactive lymphocytes were observed in ~15% of the children with equal distribution across the groups (P=0.367).

Genetic characteristics

Although heterozygous carriers of the HbS gene (HbAS) have 60-90% protection against acute malaria (Raper 1955), and reduced all-cause childhood mortality (Aidoo, et al 2002), HbAS was not significantly different (Table 1). Similarly, while G6PD deficiency confers a selective advantage against malaria (Bienzle, et al 1979), G6PD deficiency was also comparable (Table 1). Consistent with blood group O providing protection against severe malaria (Pathirana, et al 2005, Rowe, et al 2007), O⁺ individuals were over-represented in the UM, however, the across group difference was not significant (Table 1).

Haemozoin-containing neutrophils and monocytes

Since we recently demonstrated that elevated PCM was associated with increased risk of SMA (Awandare, et al 2007), presence of PCM and PCN were examined. PCN was substantially lower than PCM in all categories (Table 1). PCM prevalence and PCM/μL increased with anaemia severity, while PCN proportions and PCN/μL were similar across the groups (Table 1).

History of present illness and physical exam findings

Clinical characteristics of the study participants are shown in Table II. Though history of fever was frequent (>90%), recent febrile episodes (1-3 days) were more common in mild anaemia (78% UM, 59% non-SMA and 39% SMA, P<0.001), whereas prolonged fever (4-14 days) was coupled with enhanced anaemia (13% UM, 30% non-SMA and 54% SMA, P<0.001). Emesis was similar across groups, however, protracted emesis (4-14 days) was associated with more profound anaemia (9% UM, 14% non-SMA and 25% SMA, P=0.025). History of generalized pallor, respiratory distress, palmar and conjunctival pallor, splenomegaly and scleral icterus significantly differed and increased with anaemia severity. Visual nutritional status characterized as “fair or poor” upon physical examination was commonly observed and progressively increased with anaemia severity. Clinical assessment of height-for-age (stunting), weight-for-height (wasting), weight-for-age (underweight), head circumference-for-age, MUAC-for-age and MUAC-for-height Z scores (i.e., ≤ −2, ≤ −3 or ≤−4) revealed only one significant across group finding: a higher proportion of MUAC-for-height Z scores ≤ -2 in SMA (46%) vs. non-SMA (26%) and UM (23%) (P=0.004).

Table II.

Clinical characteristics

		UM	Non-SMA	SMA	P	n
Patient History	Fever	21 (91)	185 (92)	128 (97)	0.202	355
	Emesis	9 (39)	118 (58)	82 (62)	0.118	355
	Poor feeding	16 (70)	144 (72)	104 (80)	0.247	355
	Oedema	0 (0)	1(1)	1 (1)	0.888	355
	Jaundice	0 (0)	3 (1)	4 (3)	0.482	355
	Difficulty breathing	2 (9)	14 (7)	12 (9)	0.779	355
	Cough	6 (26)	72 (36)	43 (33)	0.573	355
	Diarrhaea	8 (35)	86 (43)	52 (39)	0.658	355
	Generalized pallor	1 (4)	27 (13)	48 (36)	<0.001	355
	Convulsions	0 (0)	4 (2)	6 (4)	0.267	355
	Unresponsiveness	0 (0)	2 (1)	0 (0)	0.463	355
Physical Exam	Axillary temperature (°C) a	37.6 (2.5)	37.5 (1.9)	37.3 (1.4)	0.397	355
	Axillary temperature >37.5°C	13 (56)	103 (51)	58 (44)	0.335	355
	Heart rate a	124 (17)	124 (20)	131 (16)	0.362	350
	Tachycardia b	2 (9)	13 (7)	7 (5)	0.793	350
	Hypotension b	3 (33)	10 (18)	24 (37)	0.065	128
	Respiratory rate a	32 (14)	32 (13)	32 (12)	0.755	337
	Tachypnea b	3 (14)	17 (9)	13 (10)	0.740	336
	Respiratory distress c	0 (0)	1 (1)	9 (7)	0.002	355
	Palmar pallor	0 (0)	40 (20)	82 (63)	<0.001	354
	Conjunctival pallor	1 (4)	84 (42)	101 (76)	<0.001	355
	Dehydration d	0 (0)	8 (4)	8 (6)	0.372	355
	Splenomegaly	0 (0)	27 (13)	40 (30)	<0.001	354
	Hepatomegaly	0 (0)	2 (1)	1 (1)	0.877	354
	Lymphadenopathy	0 (0)	10 (5)	7 (5)	0.532	354
	Scleral icterus	0 (0)	0 (0)	4 (3)	0.032	355
	Fair or poor nutritional statuse	11 (48)	108 (54)	100 (76)	<0.001	355

^aMedian (interquartile range) of axillary temperature, heart rate, and respiratory rate is reported. For all other clinical signs or symptoms, the number (%) of patients is reported.

^bTachycardia, hypotension, and tachypnea were age-adjusted according to standard reference values.

^cRespiratory distress was defined as the presence of any of the following signs: alar flaring; chestretraction; use of accessory muscles during respiration; or abnormally deep, acidotic breathing.

^dDehydration was defined as presence of sunken eyes or fontanelles, dry mucosa, or decreasedskin turgor.

^eNutritional status of the children was determined by assessment of visible malnutrition, coded as “good” vs. “fair or poor”.

Predictors of SMA

To identify predictors of SMA (Hb<6.0 g/dL), logistic regression models were constructed. Cluster-specific regressions identified the following statistically significant variables: age; reticulocyte count; platelet count; prevalence of PCM; history of pallor; fever duration; conjunctival and palmar pallor; and splenomegaly (Table III). Among these variables, age, reticulocyte count, prevalence of PCM, history of pallor, and conjunctival and palmar pallor emerged as significant predictors of SMA in the second-step (final) regression analysis (Table III).

Table III.

Binomial logistic regression analyses examining clinical predictors of SMA

	SMA (Hb<6.0 g/dL)				SMA (Hb<5.0 g/dL)
	Predictor	OR	95%CI	P	Predictor	OR	95%CI	P
Demographic	Age	0.94	0.90-0.98	0.006	Age	0.96	0.92-1.01	0.093
Laboratory	Reticulocyte count	1.08	1.01-1.17	0.026	MCV	1.04	1.01-1.07	0.009
	Platelet count	0.99	0.99-1.00	0.201	Platelet count	0.99	0.73-1.54	0.492
	PCM	4.37	2.39-7.97	<0.001	PCM	2.71	1.41-5.21	0.003
History	Pallor (history)	2.26	1.16-4.41	0.017	Pallor (history)	1.38	0.70-2.73	0.357
					Convulsions (history)	6.53	1.37-31.24	0.019
Physical Exam	Conjunctival pallor	2.16	1.10-4.22	0.025	Palmar pallor	3.71	1.97–6.97	<0.001
	Palmar pallor	4.24	2.21-8.11	<0.001	Respiratory distress	4.82	0.89-31.24	0.068
	Splenomegaly	1.88	0.92-3.83	0.081

Final regression analyses performed using the two alternative Hb cut-off values to define SMA. All variables that emerged as significant predictors of SMA in the first-step regression analyses are shown within their respective regression clusters.

Odds Ratio (OR) and 95% confidence interval (CI).

Pigment-containing monocytes (PCM).

Mean corpuscular volume (MCV).

Stratification according to the WHO definition of SMA (Hb<5.0 g/dL) yielded 67 children in the SMA group and 265 in the non-SMA group. The initial cluster-specific regressions revealed that age, MCV, platelet count, prevalence of PCM, history of pallor, history of convulsions, palmar pallor and respiratory distress were significant predictors of SMA (Table III). In the second-step (final) regression analysis, MCV, prevalence of PCM, history of convulsions and palmar pallor were significant predictors of SMA (Table III). Respiratory distress was associated with a 4.8-fold higher risk of SMA, but results did not reach statistical significance (P=0.068, Table III).

DISCUSSION

Severe malarial anaemia is the most common clinical manifestation of severe malaria in infants and young children in holoendemic P. falciparum transmission areas and accounts for the greatest amount of global malaria-related morbidity and mortality (Breman, et al 2001, WHO 2005). In a retrospective investigation of hospital admissions in Siaya District, the same hospital as the current study, severe anaemia accounted for 53% of the malaria-related deaths (Obonyo, et al 2007). Children in Siaya, as with many other holoendemic P. falciparum transmission areas, have a high rate of co-morbidities that can have a substantial impact on the development of SMA (Otieno, et al 2006). As such, it becomes difficult to discern the haematological and clinical factors characteristic of SMA versus those that arise due to other pathogens. It was therefore important to comprehensively examine clinical predictors and haematological characteristics of SMA in a large cohort of children in the absence of known co-morbidities that influence the course and outcomes of malarial anaemia.

The current study focused on SMA in infants and young children (3 to 36 mos.), a paediatric population highly susceptible to severe anaemia, as evidenced by a recent investigation in Siaya showing that SMA is found almost exclusively (i.e., 89% of the cases) in children less than three yrs. of age (Obonyo, et al 2007). This former study, one of few large cohorts investigated in a holoendemic P. falciparum transmission area, also illustrated that SMA has a 26% prevalence rate in children aged one to five years of age (Obonyo, et al 2007). As such, children selected for the present study were recruited during their first hospital contact for the treatment of paediatric illness so that they fell within this important age range. In addition, children were not eligible for enrolment if they had a prior hospitalization (for any reason) or transfusion. This strategy was aimed at reducing the potential confounding effects associated with previous illnesses, while at the same time, enriching the overall number of SMA cases presenting at hospital.

Consistent with our previous investigation in western Kenya (Ong’echa, et al 2006), anaemia severity was not associated with malaria parasite levels. For example, children with uncomplicated malaria had the highest peripheral parasitaemia, geomean parasitaemia and proportion of HDP, suggesting that parasite burden per se is not the primary factor responsible for the development of SMA. Although beyond the scope of the current manuscript, numerous investigations by our group support this finding and suggest that the immune response to P. falciparum, rather than the cumulative parasite burden, is responsible for promoting SMA in holoendemic regions (Awandare, et al 2006, Ong’echa, et al 2006, Ouma, et al 2006).

As shown in Table 1, children with malarial anaemia (i.e., non-SMA and SMA) had a lower MCV and MCH than individuals with uncomplicated malaria. Moreover, MCV was a significant predictor of SMA (Hb<5.0 g/dL) in the multivariate model. Although it is possible that iron deficiency may contribute to the current findings, these parameters were not determined since it was difficult to obtain a large enough blood volume from small, anaemic children to perform a number of desired investigations. Increasing reticulocytosis and an elevated reticulocyte count were also significantly associated with enhanced anaemia severity. However, the prevalence of children with ineffective erythropoiesis (i.e., RPI<2) was also elevated in the non-SMA group (72%) and highest in children with SMA (81%), supporting our previous results demonstrating that malarial anaemia, particularly SMA, in holoendemic regions is characterized by ineffective erythropoiesis and/or bone marrow suppression (Keller, et al 2009, Were, et al 2009).

Thrombocytopaenia is a commonly observed haematological finding in malaria-infected individuals and typically presents in the mild-to-moderate range (Ladhani, et al 2002). During a malarial infection, the lifespan of platelets is reduced through a number of defined mechanisms, including platelet adhesion to the endothelial cells and RBCs, and increased platelet clearance in the setting of hypersplenism (Kueh and Yeo 1982). Findings here show that as anaemia progressively worsens, platelet counts decrease with accompanying splenomegaly. Concomitantly, these results suggest that hypersplenism may, at least in part, contribute to thrombocytopaenia in the cohort. Consistent with previous findings (Abdalla and Pasvol 2004), however, platelet count did not emerge as a significant predictor of anaemia severity.

Studies conducted by our group (Keller, et al 2009, Were, et al 2009, Were, et al 2006) and others (Akman-Anderson, et al 2007, Casals-Pascual, et al 2006, Prato, et al 2008) support the notion that acquisition of malarial pigment (haemozoin) by leukocytes promotes dysregulation in cytokines, chemokines and effector molecules that, in turn, are associated with suppression of erythropoiesis. Moreover, from a clinical standpoint, density of haemozoin in mononuclear cells (i.e., PCM) is a better correlate of disease severity than parasite density (Awandare, et al 2007b, Lyke, et al 2003, Metzger, et al 1995). Since the lifespan of monocytes is substantially longer than that of neutrophils (Khanna-Gupta and Berliner 2008, Whitelaw 1966), prevalence of PCM can provide insight into the duration of illness. In the present study, children with SMA had a higher prevalence and density of PCM than PCN, suggesting that SMA is characterized by a longer duration of illness, or alternatively, repeated episodes of untreated subclinical malarial infections. Prevalence of PCM was also one of the strongest predictors of risk for developing SMA (at both Hb<5.0 and <6.0 g/dL, respectively) demonstrating its importance as a marker of disease severity.

In addition to presence of PCM suggesting a more protracted illness in children with SMA, the longer duration of fever, emesis and poor feeding in this group further suggests that these children have been ill for an extended period of time prior to presentation at hospital compared to children in the UM and non-SMA groups. These data are consistent with findings in Colombia showing that paediatric and adult populations with malarial anaemia have an inverse relationship between Hb levels and duration of illness (Zamora, et al 2005). Results presented here showing that parasitaemia was not correlated with anaemia, and that scleral icterus was only rarely observed, also parallel those investigations (Zamora, et al 2005). Thus, intravascular haemolysis caused by hyperparasitaemia does not likely explain the development of anaemia in the current cohort.

It is important to note that although children with SMA had a more prolonged duration of fever, the prevalence of axillary temperature (>37.5°C) was actually lower in children with SMA upon presentation at hospital. These data support findings in Burkina Faso in which afebrile P. falciparum parasitaemia contributed to a high prevalence of moderate to severe anaemia in children 6 to 23 mos. of age (Ouedraogo, et al 2008).

Anaemia may be difficult to recognize in resource-poor settings where basic laboratory investigations are not always available. Pallor has been extensively evaluated as a clinical sign of anaemia in infants and young children in malaria endemic areas (Chalco, et al 2005, Obonyo, et al 2006). Findings in the current study illustrate that the parents/guardians were readily able to identify pallor. History of pallor, along with the identification of pallor (conjunctival and palmar) by clinicians were significantly associated with enhanced anaemia severity (Hb<6.0 g/dL), while only palmar pallor was a significant predictor of SMA when using the WHO criteria of SMA (Hb<5.0 g/dL). Identification of pallor by caregivers and clinical staff therefore represents an important clinical sign that can be used to identify more profound anaemia.

Respiratory distress is a strong correlate of life-threatening malaria in non-holoendemic transmission areas (Marsh, et al 1995). In the current study, respiratory distress was strongly associated with SMA in the univariate analyses, although it was not significantly associated with SMA in the final regression models using either definition of SMA (i.e., Hb<5.0 or <6.0 g/dL). However, the nearly 5-fold increased risk of developing SMA associated with respiratory distress when using the WHO criteria of SMA demonstrates that respiratory distress is highly important and may not have reached statistical significance in our cohort due to sample size issues. The finding that tachypnea was not significantly correlated with SMA appears to suggest that the patho-physiological process underlying respiratory distress was related to metabolic acidosis, rather than shallow rapid breathing due to pulmonary edema. This interpretation is consistent with investigations showing a high prevalence of metabolic acidosis in children with severe malaria in which respiratory rate and auscultatory findings failed to predict disease severity (Marsh, et al 1995).

Convulsions emerged as a significant predictor of SMA in the final regression model, but only when using the WHO definition of SMA [i.e., Hb<5.0 g/dL (Snow, et al 1999a)]. Although a history of convulsions was reported by the childrens’ parents/guardians, convulsions were not observed by clinicians when children presented at hospital, nor were convulsions significantly correlated with fever, age or hypoglycaemia (data not presented), suggesting that children may have had benign febrile seizures prior to presentation at hospital that did not manifest as neurological sequelae upon physical examination.

Micro- and macro-nutrient deficiencies are common in sub-Saharan Africa and likely impair the host immune response to parasitic diseases, possibly preventing the development of semi-immunity to severe malaria that occurs in endemic transmission areas. However, previous studies have shown that malnourished children are partially protected against malaria (McGregor 1982). This counterintuitive finding underscores the complexity of the interaction between the host immune system and the malaria parasite, and may be explained by postulating that a robust parasite-driven, pro-inflammatory state elicited in a well-nourished, fully immune-competent host could be maladaptive, and paradoxically, lead to higher malaria-related morbidity. In our study, we were interested in exploring the effect of clinical nutritional markers on malarial anaemia. Although the univariate analyses showed a trend towards a higher prevalence of markers of malnutrition in the SMA group, nutritional parameters did not emerge as significant predictors of SMA in the final regression model, a finding that confirms previous observations by our group in this region (Ong’echa, et al 2006). One explanation for this finding is that the effect of iron deficiency outweighs that of other nutritional deficiencies and adversely affects erythropoiesis. Thus, iron deficiency may mediate the correlation observed between malnutrition and malarial anaemia. Although we did not directly test for iron deficiency, a limitation imposed by the need to minimize the risk of iatrogenic anaemia, the significant correlation between a low MCV and SMA in the final regression model supports previous reports where stunted patients with malaria were more likely to have severe anaemia and a high serum transferrin receptor concentration, a sensitive marker of iron deficiency (Verhoef, et al 2002). However, this postulate needs to be validated since a recent study in preschool children in Malawi showed that iron deficiency was highest in community and hospital control patients relative to those with SMA (Calis, et al 2008). A previous study in western Kenya, conducted in children in the same age range, showed that stunting was more common in children with SMA (Friedman, et al 2005). This apparent discrepancy with findings presented here may be related to the fact that the confounding effects of HIV-related malnutrition were excluded in our population.

A previous study exploring the risk factors associated with malarial anaemia in Zambian children aged 6 to 119 mos. found that hookworm infestation was present in only 1.2% of the study participants (Mulenga, et al 2005). Since the highest prevalence of hookworm in African children typically occurs between six to ten years of age (Diemert, et al 2008), an even lower prevalence of hookworm infections may be expected in children less than 36 mos. of age examined here. Comprehensive hookworm infestation data were not available in the current study due to difficulty in obtaining stool samples, particularly in very ill children. As such, the influence of hookworm infections on anaemia outcomes presented here cannot be definitively ruled out.

In summary, we have comprehensively characterized the demographic, clinical, and haematological parameters associated with SMA and identified a number of predictors of SMA in pre-school children from western Kenya. Significant predictors of SMA when using a definition of Hb<6.0 g/dL, include: age; reticulocyte count; PCM; history of pallor and conjunctival and palmar pallor, whereas significant predictors of SMA when using a definition of Hb<5.0 g/dL, include: MCV; PCM; history of convulsions; and palmar pallor. Thus, it appears that there are both unique and overlapping clinical predictors of SMA depending on the definition of anaemia severity employed. Nevertheless, it is clear that both pallor and intra-monocytic hemozoin deposition can be used to identify those children that suffer from SMA. Results presented here have direct clinical implications since identification of the clinical predictors discovered in the current study may aid in the prompt recognition of children that require rapid and appropriate therapeutic management.