Bacteremia and Malaria in Tanzanian Children Hospitalized for Acute Febrile Illness

Abstract

We recorded the reason for presentation to a rural hospital in an area endemic for malaria in 909 children between January 2006 and March 2009. Blood smears were examined for Plasmodium falciparum parasites, and blood spots dried on filter paper were prepared for 464 children. A PCR assay utilizing the stored blood spots was developed for Streptococcus pneumoniae (lytA) and Haemophilus influenzae (pal). Malaria was present in 299 children whose blood was tested by polymerase chain reaction (PCR); 19 had lytA and 15 had pal. The overall prevalence of lytA was 25 of the 464 children, while that of pal was 18 children. Fever was present in 369 children of whom 19 had lytA DNA while 11 had pal DNA detected. Of the 95 afebrile children, six had lytA and seven pal. We conclude that there are no clinical features that distinguish malaria alone from bacteremia alone or the presence of both infections.

INTRODUCTION

Malaria and acute lower respiratory infections (ALRIs) combined are responsible for greater than one third of the pediatric deaths in Africa [1] In resource-poor settings, the inability to distinguish complicated malaria from lower respiratory infection in a febrile child may result in inadequate therapy and increased mortality. In children less than 5 years old, WHO has estimated that ALRI is responsible for more childhood deaths than malaria, measles and AIDS combined. The bacteria responsible for ALRI include non-typhoidal Salmonella, Haemophilus influenzae and Streptococcus pneumoniae [2]. However, only type b H. influenzae and certain pneumococcal serotypes are vaccine preventable.

In sub-Saharan African countries with a high prevalence of malaria, children are often brought to the hospital with fever and respiratory symptoms. The approach to such children involves an examination of a peripheral blood smear for malarial parasites, and if present, administration of artemether-lumefantrine or injectable quinine. In children who do not have malaria but with clinical features of an ALRI, antibiotics are administered.

METHODS

Research setting

The study was conducted at Morogoro Regional Hospital in Tanzania, a referral hospital for 14 district-level hospitals, 26 health centers and 240 dispensaries. Morogoro is located 193 kilometers west of Dar es Salaam. During the period of the enrollment, January of 2006 to December of 2009, the monthly pediatric census at the Morogo Regional Hospital ranged from 1901 to 3418 children.

Recruitment of patients

In all, 909 children from the newborn period to 60 months were enrolled: the mean age was 24.4 ± 19 months, with a median age of 20 months.

On admission, the temperature of the tympanic membrane was recorded as well as the respiratory rate, oxygen saturation in room air, chest ‘in-drawing’, vomiting, cough, diarrhea and convulsions. Data were collected from two cohorts between January 2006 and April 2007 and from May 2007 to March 2009. There were no statistically significant differences between the two groups with regard to the chief complaint, diagnosis at admission, degree of parasitemia, presenting symptoms or the presence ALRI. Fever was defined as a tympanic temperature greater than or equal to 37.5°C. The presence of ALRI was defined by the WHO criteria: respiratory rate > 60 breaths/minute in infants less than 2 months old, >50 breaths/min in children 2–12 months and > 40 breaths/min in children 12–60 months old with chest retractions or hypoxia (O₂ saturation < 93% by pulse oximetry in room air [3, 4]).

Laboratory evaluation

Blood from 464 children with an acute febrile illness was tested for H. influenzae DNA, using primers for the pal gene and for the lytA gene of S. pneumoniae. Blood was collected on a Whatman 903 card used for newborn metabolic screening, allowed to dry and placed in a plastic envelope and stored at room temperature. Specimen tracking was performed with a bar code using the Legacy or DataFax data system. At the same time, a thin and thick smear was prepared and examined by laboratory personnel trained to recognize Plasmodium falciparum. The Giemsa-stained thick peripheral smears were scored as positive or negative for parasites after examination of approximately 10 000 erythrocytes and recording the positive samples an estimated number of organisms per high power field.

Statistical analysis

Simple summary statistics are presented and associations are tested using Fisher’s exact tests, and 2-sided p-values are reported at the 0.05 significance level. SAS Version 9.2 (Cary, NC, USA) was used for all analyses. The study was reviewed and approved by Western Institutional Review Board (WIRB) in Seattle, WA, and the Medical Research Coordinating Committee of the National Institute for Medical Research in Tanzania.

MOLECULAR METHODS

Assay development

To determine the sensitivity of detection of DNA specific for H. influenzae and S. pneumoniae, multiple isolates of each of these species were cultured on chocolate agar plates and the DNA purified from colonies using the Qiagen DNA Easy Kit. H. influenzae and S. pneumoniae isolated from normally sterile site of infants in Gambia, Kenya, South Africa, Papua New Guinea, Bangladesh and the USA, obtained from the collection of one of the authors (A.L.S.), were tested. The primary reference H. influenzae is strain R1834, a nontypeable H. influenzae cultured from a lung aspirate of a child in Papua New Guinea (A.L.S. collection). Strain R3751 is a capsule type 14 S. pneumoniae (NCTC 11902) [5] and was the reference pneumococcus. H. haemolyticus ATCC# 10014 was obtained from the American type Culture Collection as was H. parainfluenzae (ATCC # 9796) and serotypes a (ATCC # 9006), b (ATCC # 9795), c (ATCC # 9007), d (ATCC # 9008), e (ATCC # 8142) and f (ATCC #9833).

The blood spot was divided into two equal parts and the DNA was extracted using an 8 µL aliquot of the 10 µL extracted sample; the limit of detection with H. influenzae was 45 × 10⁻¹² grams of DNA/µL or 45 pgm/µl. With the pneumococcal primer lytA, the detection of 3.4 × 10⁻¹⁵ gm/µl of the reference pneumococcal DNA was achieved. For comparison, an Escherichia coli strain 82/r cell is calculated to contain 8 × 10⁻¹⁵ gm of DNA [6].

Since PCR detects bacterial DNA, the amount that correlates with viable cell density is dependent on the growth phase of the organism.

PCR assay

Details of the PCR assay are available on request to Professor A. Smith. Only samples with concurrence of the duplicates and containing β-globin were tabulated.

RESULTS

H. influenzae cultured in vitro but isolated from normally sterile sites from children living in developing countries, as described above, were tested. The pal primers did not detect DNA of H. haemolyticus or H. parainfluenzae when added to blood at 2 × 10⁸ CFU/ml blood, but did detect H. influenzae serotypes a, b, c, d, e and 16 nontypeable isolates. We stored 50 samples at room temperature (25 ± 4°C) for 26 months on filter paper; all contained amplicons for β-globin, pal and lytA DNA after storage. The limit of detection of viable H. influenzae was 830 CFU/ml blood, whereas with the lytA primer, the limit of detection of pneumococci was 10 CFU/ml blood.

Table 1 depicts the demographic data of all 909 subjects admitted to the hospital and the subset of 464 children whose blood was examined for malaria parasites and tested for pal and lytA. Although 79% of the children had a tympanic temperature >37.5°C, only 39.3% (n = 123) of these had respiratory symptoms which met the criteria for the diagnosis of ALRI.

Table 1.

Clinical details of admission events

	All events (N = 909)		PCR’d events (N = 464)
	N	%	N	%
Admission reason: malaria	847	93.2	440	94.8
Positive maleria blood smear	442	48.6	299	64.4
Blood smear maleria parasites: 0	467	51.4	165	35.6
1–999	221	24.3	142	30.6
1000–4999	120	13.2	90	19.4
≥5000	101	11.1	67	14.4
Vomiting	227	25.0	128	27.6
Convulsions	228	25.1	116	25.0
Cough	282	31.0	150	32.3
Diarrhea	32	3.5	17	3.7
Fatigue	56	6.2	30	6.5
Abdominal discomfort	30	3.3	18	3.9
Reduced appetite	27	3.0	18	3.9
Tachypneaa	356	39.2	149	32.1
Chest Indrawing	141	15.5	73	15.7
Hypoxiab	120	13.2	42	9.1
WHO-defined ALRIc	150	16.5	61	13.2
Maximum body temperature >37.5°C	719	79.1	369	79.5
WHO ALRI + Fever (>37.5°C)	123	13.5	46	9.9
Outcome
Unknown	118	13.0	79	17.0
Survive	736	81.0	363	78.2
Died	55	6.1	22	4.7

^aRespiratory rate >50 under 12 months of age; respiratory rate >40 12 months and older.

^bO2 sat <93%.

^cEvidence of tachypnea with chest indrawing OR hypoxia.

N = number of subjects.

Among 464 hospitalized children, 9.3% had PCR bacteremia based on the detection of bacterial DNA: 18 children with H. influenzae and 25 with S. pneumonia (Table 2). Of 61 (13.2%) admitted children who met the current criteria for WHO-defined ALRI, 4 (6.6%) had PCR detectable bacteremia. Of the 53 (11.4%) children with WHO-defined ALRI and fever, 5.7% had bacteremia. Prevalence of either bacterium was 12.5% in those without fever or WHO-defined ALRI and 9.1% in patients with fever alone. Thus, the presence of fever does not predict bacteremia in children with ALRI. These differing incidence rates were not statistically significant (p = 0.43). There was no significant difference in PCR detectable bacteremia in those with and without WHO ALRI symptoms (6.6% vs. 9.7%, p = 0.49).The small numbers preclude accurate determination of whether the presence of malaria increased the risk of bacterial infection in those patients: 3/40 with malaria vs. 1/21 without malaria (p = 1.0).

Table 2.

PCR-detectable bacteremia by symptom and malaria subgroups

	N	PCR positive
		H. influenzae N (%)	S. pneumoniae N (%)
Overall	464	18 (3.9%)	25 (5.4%)
Fever
>37.5°C	369	11 (3.0%)	19 (5.2%)
≤37.5°C	95	7 (7.4%)	6 (6.3%)
WHO ALRIa	61	1 (1.6)	3 (4.9%)
Without fever (≤37.5°C)	8	1 (12.5%)	0 (0%)
With fever (>37.5°C)	53	0 (%)	3 (5.7%)
Malaria
Positive	299	15 (5.0%)	19 (6.4%)
Negative	165	3 (1.8%)	6 (3.6%)
Malaria (+) w/ WHO ALRI	40	1 (2.5%)	2 (5.0%)

^aWorld Health Organization (WHO) definition of Acute Lower Respiratory Infection (ALRI) by evidence of tachypnea with chest indrawing OR hypoxia.

N = number of subjects.

At the Morogoro Regional Hospital in Morogoro, Tanzania, the site of the present study, the prevalence of P. falciparum parasitemia in mothers at delivery was 11.5% [9]. Rachas et al. [10] found that placental malaria is associated with an increase in non-malaria infection in infants during the first 18 months of life. In Kenyan children aged 3 months to 13 years, the bacteremia incidence rate ratio associated with malaria parasitemia was 6.69 (95% CI, 1.31–34.3), with 62% of bacteremic children having malaria [11]. Although ‘malaria with fever’ was the presenting complaint in the majority of our subjects, it was not corroborated by tympanic temperatures greater 37.5°C, present in only 79% of our subjects. We also found that fever was a poor predictor of bacteremia with H. influenzae or S. pneumoniae: only 369 children with fever were tested by PCR and only 30 had bacteremia with S. pneumoniae or H. influenzae (Table 2). In a recent study of febrile children in Teule Tanzania, blood cultures yielded bacteria in 336 of 3639 subjects, and of the positive cultures, nontyphoidal Salmonella was recovered in 162 subjects [12]. In contrast, in a study in Malawi, 287 of the 299 bacteremic children were febrile [13], and 15 of 317 children had a positive blood culture. In a Kenyan study of bacteremia in children with malaria, where fever was defined as axillary temperature ≥37.5°C, 44.7% of the children with fever were not bacteremic, while only 56% of the bacteremic children were febrile [11]. Teng et al. [14] conducted a meta-analysis of the literature from developing countries and found that the mothers’ perception of fever had a sensitivity of 89.2% and a specificity of 50%, leading them to conclude that the maternal reporting was more useful to exclude fever [14].

We found hypoxia, an arterial oxygen saturation <93% in room air in only 13.2% of the children enrolled (Table 1). In 359 ill children aged between one and 60 months admitted to Gorka Hospital in New Guinea, the median SpO₂ was 86%, with a range of 76% to 93%, including 10 children with sepsis having a median value of 79% (range 57%–94%) [14]. Only five children with malaria were in the study and all had an SpO₂ in the normal range of 92% to 96% [15].

The simultaneous bacteremia and parasitemia does not allow the attribution of symptoms, such as fever or respiratory symptoms, to one or the other or to the combination of pathogens.

FUNDING

This work was supported by T32 HD07233 from the National Institute of Child Health and Human Development: The Seattle Children’s Research Institute Seed Fund; the Foundation for NIH, The Gates Foundation Grand Challenges in Global Health (grant 1364) to PED and the Intramural Research Program, NIAID, NIH (PED).