A population-based study of the prevalence and risk-factors of low-grade Plasmodium falciparum malaria infection in children aged 0-15 years old in northern Tanzania

Abstract

Objectives:

Tanzania experiences significant malaria-related morbidity and mortality, but accurate data are scarce. We update the data on patterns of low-grade Plasmodium falciparum malaria infection among children in northern Tanzania.

Methods:

P. falciparum malaria prevalence (pfPR) was assessed in a representative sample of 819 children enrolled in 94 villages in northern Tanzania between October 2015 and August 2016, using a complex survey design. Individual- and household-level risk-factors for pfPR were elicited using structured questionnaires. pfPR was assessed using rapid diagnostic tests (RDTs) and thick film microscopy (TFM). Associations with pfPR, based on RDT, were assessed using adjusted odds ratios (aOR) and confidence intervals (CI) from weighted survey logistic regression models.

Results:

pfPR was 39.5% (95% CI: 31.5, 47.5) by RDT and 33.4% (26.0, 40.6) by TFM. pfPR by RDT was inversely associated with higher-education parents, especially mothers (5-7 years of education: aOR 0.55; 95% CI: 0.31, 0.96, senior secondary education: aOR 0.10; 95% CI: 0.02, 0.55), living in a house near the main road (aOR 0.34; 95% CI: 0.15, 0.76), in a larger household (2 rooms: aOR 0.40; 95% CI: 0.21, 0.79, more than 2 rooms OR 0.35; 95% CI: 0.20, 0.62). Keeping a dog near or inside the house was positively associated with pfPR (aOR 2.01; 95% CI: 1.26, 3.21). pfPR was not associated with bed-net use or indoor residual spraying.

Conclusions:

Nearly 40% of children in northern Tanzania had low-grade malaria antigenemia. Higher parental education and household metrics but not mosquito bed-net use, were inversely associated with pfPR.

Introduction

Tanzania has the third largest population living in areas of stable malaria transmission in Africa (1), of which 72% live in areas where transmission is hyper- or holo-endemic. About 17.2% live south of Lake Victoria where warm temperatures (24-32⁰C) and high rainfall (600-930 mm/year) are conducive to mosquito breeding and efficient completion of the malaria parasite life-cycle (2). Appropriately, the Tanzanian Government has prioritized malaria control as an important public health goal and, working with various partners, has implemented programs to scale up interventions to suppress malaria prevalence and significantly reduce morbidity and mortality(3). As the malaria profile in northern Tanzania approaches that compatible with malaria elimination goals(4), there is need to monitor the burden of asymptomatic infections. Asymptomatic infections are in important reservoir for new infections (5, 6) contribute to transmission (7) and asymptomatically infected people are mobile (8, 9), they can establish foci of transmission far away from their usual home (10). To inform national and sub-national malaria policies to achieve this goal requires timely and accurate statistics about the malaria patterns, particularly low-grade infections that contribute to new infections.

In addition to direct malaria-related morbidity and mortality, the high burden of malaria is also associated with high incidence of endemic Burkitt’s lymphoma (eBL). The eBL incidence is estimated to vary between 1.4 to 22 per 100,000 children in northern Tanzania (11). This is 5.6-88-fold higher than the incidence of sporadic BL in United States (12), which is not attributed to the impact of malaria). eBL is an aggressive B-cell lymphoma associated geographically with Plasmodium falciparum (Pf) malaria infection (11, 13). This association was recently confirmed by showing that the sickle cell trait, which is protective against severe malaria also protects against eBL (14). Consistent with this, geographic patterns of eBL mirror those of malaria, with particularly high incidence in areas south and east of Lake Victoria (11, 13). The patterns of eBL in this region between 2000 and 2009 showed significant regional variation suggesting that the eBL incidence in Mara Region was twice that in Mwanza Region (11). Additional sub-regional variation was noted at the district level showing a 10-fold difference in the incidence of eBL in Magu District in Mwanza Region versus that reported in Tarime District in Mara Region (11). When the frequency of cases over time was considered, there was a decline in the incidence of eBL between 2000 and 2009 (11), but this decline preceded the introduction of major malaria prevention activities in 2005(15). Although population data about malaria in northern Tanzania are available, they do not cover the entire region (16–18), are based on surveys of children attending school (19, 20), or on children in a restricted age range (16, 18, 21), making them insufficient to evaluate the role of malaria in the observed declining eBL incidence (11).

Therefore to obtain more recent data about malaria among children in order to inform policy and also to generate baseline data for eBL studies, we investigated the prevalence and risk factors for asymptomatic malaria among eBL-age children in northern Tanzania, where we are simultaneously conducting a case-control study of eBL, the EpideMiology of Burkitt Lymphoma in East African Children and Minors (EMBLEM) (22).

Methods

Geographic location and study population

Between October 2015 and August 2016, children aged 0-15 years were enrolled in areas encompassing the original Mwanza Region, which was divided in 2012 into three new regions named Mwanza, Geita, and Simiyu regions, and Mara Region of northern Tanzania (Figure 1a). This area lies on a plateau about 1000 m above sea level, between the eastern and western African Rift valley. The area is drained by seasonal rivers that flow into Lake Victoria. These geographical features, in addition to the short heavy rains interspersed by long dry seasons, location in the rain shadow of Mount Kilimanjaro, and sparse drought-resistant shrubs, influence the regional ecology.

Figure 1.

(A) Map showing the study area displayed as a zoom out with regional, district, and enumerations area (EA) boundaries and geographical features, such as lakes and rivers delineated. The EAs are colored based on the selection strata, i.e. proximity to water and population density (see Methods). Shaded circles indicate six selected EAs where enrollment of matched population children (MPC) was not done. The locator map shows the study area in Tanzania. (B) The distribution of Plasmodium falciparum prevalence (pfPR) by region of residence and separately by proximity of areas within each region to the shores of Lake Victoria (< 10 miles i.e. < 16 km versus ≥ 10 miles away). The bar graph indicates pfPR by proximity to Lake Victoria for each region.

A total of 100 villages were randomly selected using a stratified multistage sampling design, as previously described (23). Briefly, the primary sampling unit was the “ward”, which is equivalent to a census enumeration area (EA) in Tanzania (Figure 2). EAs were selected from a list of EAs obtained from the Tanzania National Bureau of Statistics, stratified by ‘low population-density’, ‘high population-density’, ‘near surface water’, and ‘far from surface water’, with the probability of selection proportional to size. The EA strata were defined as high population-density, a surrogate for more urbanized areas, if they had a population count of ≥6553 children (the average count for EAs in the study area) aged 0-15 years, otherwise they were classified as ‘low population-density’, a surrogate for more rural areas. EAs were defined as near surface water (a lake, all season river or swamp) if any part of their boundary was within 500 meters of surface water; otherwise they were classified as ‘far from water’. Malaria transmission was expected to vary across these strata (23). Because EAs typically consist of 2-20 villages, one village was randomly selected per EA, and households in that village enumerated to generate a list from which 22-25 households with eligible children were randomly selected. Enrolled children were healthy, 0-15 years old at enrollment, had resided in the study area for at least 4 months(23), and were selected to achieve an age-range and sex distribution similar to that of eBL cases in the study area (11).

Figure 2.

Flowchart showing the sampling design used to identify and enroll 819 healthy children aged 0-15 years old in villages across northern Tanzania between October 2015 and August 2016.

Participant enrollment and malaria testing

Experienced field teams enrolled the children at their home after obtaining consent. Individual- and household-level risk-factor information was elicited using structured questionnaires, and venous blood samples for research (10 ml) and clinical tests (4 ml) were collected in EDTA tubes. Clinical samples were immediately tested for malaria using an antigen-antibody capture rapid diagonistc test (RDT) kit (CareStart™ MALARIA HRP2 (Pf), ACCESSBIO, Somerset, New Jersey (24) and for HIV using commercial RDT kits, per the National HIV Testing Guidelines. Thick film microscopy (TFM) was performed to visualize asexual malaria parasite forms in thick blood smear slides stained with 10% Giemsa for 10 minutes. Visualized parasites were counted against 200 white blood cells (WBCs), standardized to measured WBC count/µL, and the results from positive slides summarized as geometric mean parasite density (GMPD)/µL. Thin film slides were used to identify malaria species, and as >98% of the parasites were P falciparum, this species is assumed hereafter. We report P. falciparum prevalence (pfPR) based on RDT because RDT-positivity captures both parasitemia (visualized parasites in blood) as well as antigenemia in patients who have cleared parasites but still have catabolized parasite antigens in circulation (25). Thus, a positive RDT result in a child who was TFM negative was classified as malaria positive and not as a false-positive result. Children with symptomatic malaria (defined as fever (temperature ≥37.5 ºC) and parasite density >2500 parasites/µL) were not eligible for enrollment; when encountered they were referred for treatment at a local clinic per National Treatment Guidelines. Research samples were transported in cold boxes to local EMBLEM laboratories within two hours of collection where they were centrifuged at 1300 g for 15 minutes to separate them into plasma, buffy coat, and red blood cell fractions for storage in barcoded cryovials at −80º C.

Data management and quality control

Questionnaires and laboratory forms were reviewed for completeness, clarity and logical consistency before computerization using DataFax, which utilizes intelligent character recognition to perform data entry and minimizes data entry errors. Computerized data were reviewed, and inconsistencies were corrected before generating analysis files.

Statistical analysis

The reported results are weighted estimates that account for the differential probabilities of children being selected into the study, as previously described (23) and are thus representative of children in this region. The variance estimation accounted for sampling weights and the clustering of children within villages. The primary and tertiary weights were rescaled to match the estimated population census data, but no rescaling was necessary for secondary sampling weights because they matched the data from the Tanzania National Bureau of Statistics. The final weights were trimmed by replacing the value of the weights in the highest 1.5^th percentile of the final weight distribution with the value of the final weight at the 98.5^th percentile to minimize undue influence by children with extreme weights.

The distribution of demographic, geographical, parental, malaria prevention, and household characteristics were examined descriptively. The pfPR (based on RDT) and GMPD were evaluated by age group, season, and stratification variables. Odds ratios (ORs) and Wald-type 95% confidence intervals (95% CIs) of associations between pfPR and age group, region, season of enrollment, and individual- and household-level risk-factors were evaluated using weighted logistic regression models. Multivariate analysis was performed to determine independent association with pfPR for all covariates associated with pfPR (p<0.10) in the univariate models. Two-sided p<0.05 values were considered statistically significant. Statistical analyses were performed using SAS (version 9.4, Cary, North Carolina).

Ethical approval and participant consent

Ethical approval to conduct the study was given by the National Institute of Medical Research in Tanzania Medical Research Coordinating Committee and the National Cancer Institute Special Studies Institutional Review Board (Protocol #: 10-C-N133). Written informed consent to participate was obtained from guardians of eligible children (typically a parent) and assent was additionally obtained from children aged 7 years or older.

Results

We enrolled 819 children (99.4%, 819/824 of those invited) from 2511 households (16% of n=15686 households with eligible children) in 94 villages (Figure 2) for a weighted sample of 3,747,461 children 0-15 years old in the study area. Enrollment was not done in 6 selected villages (Figure 1) due to logistical reasons.

Characteristics of study population

The weighted distribution of characteristics of the study population are presented in Table 1 and Supplementary Table 1. Most children (44.0%, n=405/819) were aged 6-10 years old, 53.0% (432/819) were males, 32.9% (418/819) resided in a low population density village, and 80.2% (408/819) lived in a village located near surface water. Geographically, 48.8% (291/819) were from Mwanza and Simiyu Region (only one village was selected in Simiyu Region, so Simiyu and Mwanza were treated as one region), 32.7% (437/819) were from Mara Region and 18.5% (91/819) were from Geita Region. Most children (69.8%, 413/819) lived in villages located less than 10 miles (i.e.< 16.1 km) from the lakeshore, and 59.7% (424/819) were enrolled during the wet season. Most parents reported receiving 5-7 years of education (mothers: 75.7%; 615/819 and fathers: 85.4%;702/819), consistent with the Tanzanian Government policy to offer free primary education, but the percentage reporting higher-level education was small. Most parents reported their occupation was peasant farming (mothers: 89.7%; 724/819 and fathers: 84.0%; 669/819) and 88.0% of mothers (730/819) reported an income of less than 129,545 Tanzanian Shillings per month, equivalent to US$ 1.90 per day which is the World Bank’s threshold for poverty.

Table 1.

Weighted characteristics of 819 healthy children aged 0-15 years old in northern Tanzania, surveyed between October 2015 and August 2016.

Characteristic	N = 819*	Weighted%† (95% CI)
Demographics
Age, years
0-5	243	35.0 (31.2, 38.9)
6-10	405	44.0 (39.6, 48.5)
11-15	171	21.0 (17.1, 24.9)
Sex
Female	387	47.0 (42.8, 51.2)
Male	432	53.0 (48.8, 57.2)
Stratification variables
Proximity to water
Far (> 500 m)	411	19.8 (14.6, 25.1)
Near (≤ 500 m)	408	80.2 (74.9, 85.4)
Population density of children 0-15 years
Low (< 6553)	418	32.9 (25.4, 40.4)
High (≥ 6553)	401	67.1 (59.6, 74.6)
Geographical variables
Region
Mwanza‡	382	67.3 (54.0, 80.6)
Simiyu and Mwanza	291	48.8 (33.5, 64.1)
Geita	91	18.5 (7.5, 29.5)
Mara	437	32.7 (19.4, 46.0)
Proximity to Lake Victoria
<10 miles (<16.1 km)	413	69.8 (57.8, 81.9)
≥10 miles (≥16.1 km)	406	30.2 (18.1, 42.2)
Season of enrollment§
Dry	395	40.3 (25.8, 54.8)
Wet	424	59.7 (45.2, 74.2)
Parental characteristics
Mother’s education
Up to primary 4	201	23.9 (18.8, 29.0)
Primary 5-7	570	71.0 (66.5, 75.5)
≥ Senior secondary school	45	4.7 (2.6, 6.9)
Father’s education
Up to standard 4	111	14.0 (9.5, 18.5)
Standard 5-7	619	75.8 (71.5, 80.2)
≥ Senior secondary school	83	9.6 (5.8, 13.4)
Mother’s occupation
Trader/Sales	62	6.3 (3.8, 8.7)
Peasant farmer	724	89.7 (85.6, 93.9)
Manual laborer	30	3.7 (1.6, 5.7)
Father’s occupation
Trader/Sales	100	10.6 (6.5, 14.8)
Peasant farmer	669	84.0 (78.6, 89.4)
Manual laborer	46	5.0 (2.3, 7.6)
Mother’s incomeǁ, Tanzanian shillings
≤ 129545	730	88.0 (84.1, 91.9)
> 129545	89	12.0 (8.1, 15.9)
Malaria prevention
Slept under mosquito net the night before
No	319	34.2 (26.3, 42.2)
Yes	497	65.4 (57.5, 73.3)
Indoor residual insecticide sprayed in house in the past year
No	551	71.1 (62.4, 79.8)
Yes	257	27.7 (19.2, 36.2)
Indoor residual insecticide spraying schedule
2010-2011	485	54.1 (39.3, 68.8)
2010-2016	334	45.9 (31.2, 60.7)
History of malaria treatment
Outpatient
Yes, past 12 months	255	30.9 (23.7, 38.2)
Yes, > 12 months	69	11.2 (7.6, 14.8)
Never	492	57.5 (50.1, 64.9)
Inpatient
Yes, past 12 months	106	9.7 (6.8, 12.5)
Yes, > 12 months	136	17.3 (12.4, 22.1)
Never	574	72.7 (67.4, 78.0)

CI = confidence interval

^*This column shows unweighted frequencies for each category of the characteristics examined.

^†This column shows the weighted percent for each characteristic. The weighted percentages may not add up to 100% for some variables due to missing data. The weighted total population was estimated to be 3,747,460.8 (coefficient of variation of the final weights= 1.18).

^‡Mwanza denotes the original Mwanza Region, which was sub-divided into new regions named Mwanza, Simiyu, and Geita. Data are shown separately for the original Mwanza Region and for the new regions created from the original Mwanza Region. Data for the new Mwanza Region and Simiyu Region are combined for statistical stability.

^§January to March and July to August were classified as dry season months, while April to June and September to December were classified as wet season months.

^ǁIncome was categorized based on the international poverty line of $1.90 per a day, which is approximately equal to 129545 Tanzanian shillings for the average 30-day monthly income.

An insecticide-treated bed net was reportedly available for personal use for 68.1% (527/819) of the children; 65.4% (497/819) of these reportedly slept under the mosquito bed net the night before the interview. Indoor residual insecticide spraying (IRS) was applied in 2010-2011 in villages where 54.1% (485/819) of the children resided, whereas 45.9% (334/819) of the children lived in areas where IRS was applied intermittently during 2010-2016. IRS in the home in the past year was reported for only 27.7% (257/819) of the children. Outpatient malaria treatment 12 months before enrollment was reported for 42.1% (324/819) of the children; 11.2% (69/819) were reported to have received such treatment more than a year ago. Similarly, inpatient malaria treatment 12 months before enrollment was reported by 27.0% (242/819) of the children; 9.7% (106/819) reported such treatment more than a year ago.

Overall and regional trends in pfPR and parasite density

Weighted pfPR was 39.5% (483/800; 95% CI: 31.5, 47.5) by RDT and 33.4% (233/791; 26.0, 40.6) by TFM (Table 2). The weighted sensitivity of RDT to detect asexual parasites was 93.4% (220/233) and specificity was 87.8% (468/558, Table 2). The pfPR in the original Mwanza Region was 40.4% (146/372), but it varied across the new sub-regions, being 55.6% (47/89) in Geita and 34.5% (99/283) in the combined Mwanza and Simiyu Regions; the pfPR was 37.7% (171/428) in Mara Region (Figure 1b). The pfPR was not statistically different for children living in villages ≤ 10 miles versus farther away from the lakeshore (Figure 1b). Malaria parasitemia was low-grade (GMPD: 769.0 parasites/µL, 95% CI: 532.3, 1111.0) and parasite density showed slight variation by region (Mara: GMPD: 1306.0 [614.9, 2773.7]; Geita: 768.0 [440.8, 1338.2]; and Mwanza: 543.8 [357.8, 826.4]).

Table 2.

Weighted results of malaria rapid diagnostic test and parasitemia using thick-film microscopy by experienced local microscopists, among healthy children 0-15 years old in northern Tanzania, surveyed between October 2015 and August 2016.

	Thick film microscopy by experienced local technicians
	Negative	Positive	Total
	Weighted % (n)	Weighted % (n)	Weighted % (n)
Rapid diagnostic test
Negative	58.47 % (n=468)	2.22 % (n=13)	60.69 % (n=481)
Positive	8.11 % (n=90)	31.21 % (n=220)	39.31 % (n=310)
Total	66.58% (n=558)	33.42 % (n=233)	100% (n=791*)
			(Nweighted=3,661,654)

^*Results do not include 28 children not successfully tested: 6 had no rapid diagnostic test (RDT) results, 9 had no thick film microscopy (TFM) results, and 13 were missing both.

Compared to TFM by experienced local technicians the weighted sensitivity of detecting parasitemia by RDT was 93.4% (n=220/233; 95% CI: 88.1%, 98.7 %) and specificity was 87.8% (468/558; 83.5%, 92.1%). Note that the sensitivity and specificity of RDT are not simple fractions, they are weighted percentages of the data.

Age and seasonal trends in pfPR and GMPD

The pfPR rose slightly with age from 37.2% (n=16/43) among children aged 0-2 years old to 43.1% (118/274) in those aged 6-8 years old and remained flat in those 9-11 and 12-15 years old (Figure 3a). Overall pfPR was not significantly different by season (wet: 41.1%; n=161/416 versus dry: 37.1%; 156/384; p=0.61). However, significant seasonal patterns in pfPR were observed in low-population density villages (wet: 45.8%; 78/205 versus dry: 21.6%; 79/205; p=<0.0001, Figure 4a), but not in high-population density villages (wet: 38.2%; 83/211 versus dry: 42.9%; 77/179; p=0.64, Figure 4b).

Figure 3.

Figure showing age-specific Plasmodium falciparum prevalence (pfPR) (A) and age-specific geometric mean parasite density (GMPD) (B) in children aged 0-15 years enrolled in northern Tanzania who were tested for malaria using the malaria rapid diagnostic test (RDT) or thick film microscopy (TFM) examination. The horizontal bars connected by a vertical line represent the 95% confidence intervals for each estimate.

Figure 4.

Figure showing Plasmodium falciparum prevalence (pfPR) (A and B) and geometric mean parasite density (GMPD) (C and D) of children aged 0-15 years enrolled in northern Tanzania in the dry or wet season stratified by low-population density (A and C) or high-population density (B and D) villages. The horizontal bars connected by a vertical line represent the 95% confidence intervals for each estimate.

Age-specific GMPD increased slightly with age and peaked at 1154.8 parasites/µL (748.6, 1781.3) in children aged 6-8 years old, then fell gradually to 470.5 parasites/µL (249.3, 887.8) in children 12-15 years (Figure 3b). The GMPD was generally lower in the dry season compared to the wet season in both low-population density (dry season: 250.9 parasites/µL [95% CI: 145.2, 433.8] versus wet: 1171.0 parasites/µL [657.9, 2084.3], Figure 4c) and high-population density villages (dry season: 453.6 parasites/µL [314.9, 653.4] versus wet: 1026.6 parasites/µL [412.4, 2555.8], Figure 4d).

Risk-factors for P. falciparum malaria infection

Tables 3 and 4 show univariate associations between pfPR and individual and household characteristics. pfPR was significantly positively associated with living in Geita (versus Mwanza: OR 3.38; 95% CI: 1.71, 6.65), having a parent working as a peasant farmer (versus trader or sales, mother: OR 3.85; 95% CI: 1.64, 9.03 or father: OR 3.36; 95% CI: 1.73, 6.54), and keeping chickens (versus not: OR 1.62; 95% CI: 0.98, 2.70), or dogs (OR 1.76; 95% CI: 1.22, 2.56) inside or near the house. pfPR was significantly lower with having a parent with 5-7 years of primary education (versus less, mother: OR 0.55; 95% CI: 0.32, 0.94 or father: OR 0.56; 0.32, 0.98) or secondary education (mother: OR 0.04; 95% CI: 0.01, 0.18 or father: OR 0.05; 0.02, 0.15), living in a house near the main road (versus far: OR 0.33; 0.14, 0.77), living in a town/city (versus village: OR 0.10; 0.02, 0.51), living in a household with piped or public tap water (versus unprotected spring/well: OR 0.34; 0.14, 0.82). Likewise, living in a house connected to the electricity grid (versus not: OR 0.18; 0.06, 0.59) or in a house with 2 rooms (OR 0.43; 0.21, 0.88) or more than 2 rooms (versus one: OR 0.38; 0.21, 0.67; p trend=0.001) were all significantly associated with decreased pfPR. pfPR was inversely associated with a history of a malaria-related fever in the past 6 months (OR 0.57; 0.34, 0.95), but was not associated with a history of a non-malaria-related fever in the past 6 months (OR 1.19; 0.70, 2.04). pfPR was not associated with bed net use the previous night (OR 0.74; 95% CI 0.40, 1.37) or indoor residual spraying in the past year (OR 1.44; 95% CI 0.79, 2.62).

Table 3.

Univariate associations of factors with asymptomatic Plasmodium falciparum malaria among healthy children 0-15 years old in northern Tanzania, surveyed between October 2015 and August 2016.

		Unadjusted
Characteristics	N=800*	Weighted %	Odds ratio (95% CI)	p†
All subjects
Demographics
Age, years				0.24
0-5	237	35.2	Ref
6-10	397	39.6	1.21 (0.71, 2.05)
≥ 11	166	46.3	1.58 (0.93, 2.70)
Sex				0.14
Female	376	36.1	Ref
Male	424	42.5	1.31 (0.92, 1.86)
Design variables
Proximity to water				0.14
Far (> 500 m)	400	30.4	Ref
Near (≤ 500 m)	400	41.7	1.64 (0.85, 3.16)
Population density of children 0-15 years				0.77
Low (< 2235)	410	37.9	Ref
High (≥ 2235)	390	40.3	1.11 (0.55, 2.22)
Geographical
Region				0.04
Mwanza‡	372	40.4
Simiyu and Mwanza	283	34.5	Ref
Geita	89	55.6	2.37 (1.13, 4.99)
Mara	428	37.7	1.15 (0.58, 2.26)
Proximity to Lake Victoria
<10 miles (<16.1 km)	402	36.9	Ref	0.27
≥10 miles (≥16.1 km)	398	45.4	1.42 (0.76, 2.67)
Season of enrollment§				0.61
Dry	384	37.1	Ref
Wet	416	41.1	1.18 (0.62, 2.26)
Parental characteristics
Mother’s education				<0.001
Up to primary 4	197	52.5	Ref
Primary 5-7	556	37.7	0.55 (0.32, 0.94)
≥ Senior secondary school	44	4.0	0.04 (0.01, 0.18)
Father’s education				<0.001
Up to primary 4	108	55.5	Ref
Primary 5-7	605	41.1	0.56 (0.32, 0.98)
≥ Senior secondary school	81	6.0	0.05 (0.02, 0.15)
Mother’s occupation				<0.01
Trader/Sales	62	16.0	Ref
Peasant farmer	707	42.2	3.85 (1.64, 9.03)
Manual laborer	28	16.4	1.03 (0.18, 5.95)
Father’s occupation				<0.001
Trader/Sales	99	18.7	Ref
Peasant farmer	652	43.6	3.36 (1.73, 6.54)
Manual laborer	45	18.5	0.99 (0.27, 3.56)
Mother’s incomeǁ, Tanzanian shillings				0.24
≤ 129,545	714	41.0	Ref
> 129,545	86	28.5	0.57 (0.22, 1.47)
Malaria prevention
Slept under mosquito net the night before				0.34
No	311	44.4	Ref
Yes	486	37.2	0.74 (0.40, 1.37)
Indoor residual insecticide sprayed in house in the past year				0.23
No	541	37.5	Ref
Yes	248	46.4	1.44 (0.79, 2.62)
Indoor residual insecticide spraying schedule				0.42
2010-2011	478	36.7	Ref
2010-2016	322	42.9	1.29 (0.68, 2.45)
History of malaria treatment
Outpatient				0.10
Yes, past 12 months	247	31.0	Ref
Yes, > 12 months	69	31.6	1.03 (0.37, 2.85)
Never	481	45.8	1.89 (0.98, 3.64)
Inpatient				0.45
Yes, past 12 months	103	35.9	Ref
Yes, > 12 months	134	34.4	0.94 (0.46, 1.91)
Never	560	41.4	1.26 (0.66, 2.40)

CI = confidence interval

^*Results are based on children with complete rapid diagnostic results. Children with missing rapid diagnostic results (n = 19) were excluded from this analysis.

^†p for heterogeneity

^‡Mwanza denotes the original Mwanza Region, which was sub-divided into new regions named Mwanza, Simiyu, and Geita. Data are shown separately for the original Mwanza Region and for the new regions created from the original Mwanza Region. Data for the new Mwanza Region and Simiyu Region are combined for statistical stability.

^§January to March and July to August were classified as dry season months, while April to June and September to December were classified as wet season months.

^ǁIncome was categorized based on the international poverty line of $1.90 per a day, which is approximately equal to 129,545 Tanzanian shillings for the average 30-day monthly income.

Table 4.

Univariate associations of additional factors with asymptomatic Plasmodium falciparum malaria among healthy children 0-15 years old in northern Tanzania, surveyed between October 2015 and August 2016.

		Unadjusted
Characteristics	N=772*	Weighted %	Odds ratio (95%CI)	p†
Home characteristics
Distance of home from main road				<0.01
Far from the main road	627	45.5	Ref
Near the main road	112	21.3	0.33 (0.14, 0.77)
In town or city	57	7.9	0.10 (0.02, 0.51)
Distance to water source, meters				0.68
< 500	135	42.3	Ref
500-999	154	35.7	0.98 (0.47, 2.03)
1,000-4,999	428	41.8	0.76 (0.36, 1.58)
≥ 5,000	80	34.0	0.70 (0.27, 1.84)
Source of drinking water				0.06
Unprotected spring/well	445	45.5	Ref
Protected spring/well	210	44.8	0.97 (0.53, 1.78)
Public tap/piped household	142	21.9	0.34 (0.14, 0.82)
Connected to electricity grid				0.01
No	726	42.9	Ref
Yes	71	12.0	0.18 (0.06, 0.59)
Number of rooms in house				<0.01
1 room	165	58.8	Ref
2 rooms	305	38.1	0.43 (0.21, 0.88)
≥ 3 rooms	327	34.4	0.37 (0.21, 0.65)
Number of people sleeping in the same room as child				0.10
1-2 people	263	39.4	Ref
3 people	283	45.1	1.26 (0.76, 2.11)
≥ 4 people	250	33.4	0.77 (0.49, 1.21)
Number of usual children residents				0.83
1-2 people	265	42.0	Ref
3 people	214	38.0	0.85 (0.43, 1.66)
≥ 4 people	313	38.1	0.85 (0.49, 1.49)
Number of usual adult residents				0.76
1-2 people	547	41.0	Ref
3 people	92	35.7	0.80 (0.39, 1.62)
≥ 4 people	158	37.4	0.86 (0.48, 1.52)
Other Malaria Prevention
Owns treated mosquito bed net				0.30
No	281	45.0	Ref
Yes	516	37.2	0.72 (0.39, 1.34)
Regularly uses mosquito insecticide sprays				0.24
No	758	40.7	Ref
Yes	37	19.1	0.34 (0.06, 2.05)
History of fevers and hospital admission
Has fever at enrollment				0.92
No	748	39.6	Ref
Yes	49	40.7	1.05 (0.40, 2.75)
Reported ≥1 fever in the last 12 months				0.14
No	112	48.6	Ref
Yes	636	37.4	0.63 (0.34, 1.17)
Reported ≥1 fever due to malaria in the past 6 months				0.03
No	244	48.5	Ref
Yes	552	35.0	0.57 (0.34, 0.95)
Reported ≥1 fever not due to malaria in the last 6 months				0.51
No	587	38.8	Ref
Yes	208	43.1	1.19 (0.70, 2.04)
Reported ≥1 hospital admission				0.81
No	503	39.2	Ref
Yes	294	40.4	1.05 (0.70, 1.56)
Animals kept near or inside house
Chicken				0.06
No	161	31.1	Ref
Yes	636	42.3	1.62 (0.98, 2.70)
Pigs				0.53
No	786	39.6	Ref
Yes	11	55.5	1.91 (0.24, 14.95)
Goats				0.97
No	444	39.7	Ref
Yes	353	39.5	0.99 (0.63, 1.55)
Sheep				0.22
No	653	38.5	Ref
Yes	144	46.9	1.41 (0.81, 2.46)
Cows				0.90
No	471	39.9	Ref
Yes	326	39.1	0.97 (0.58, 1.61)
Birds				0.26
No	699	40.8	Ref
Yes	98	31.4	0.66 (0.32, 1.37)
Dogs				<0.01
No	324	32.8	Ref
Yes	473	46.3	1.76 (1.22, 2.56)

CI = confidence interval

^*Results shown do not include children with missing rapid diagnostic test results (n = 19).

^†p for heterogeneity

In the multivariate analysis, five of the 12 variables (p<0.10) remained significantly associated with pfPR (Table 5). pfPR was significantly inversely associated with increasing level of education of either mothers’(p trend=0.01) or fathers’ (p trend=0.03) level of education and this was observed for parents with 5-7 years of primary education (mother: OR 0.55; 95% CI: 0.31, 0.96 and father: OR 0.85; 95% CI: 0.44, 1.64) or with senior secondary education (mother: OR 0.10; 95% CI: 0.02, 0.55 and father: OR 0.14; 95% CI: 0.04, 0.49) and with living in a house near the main road (OR 0.34; 95% CI: 0.15, 0.76), in a house with 2 rooms (versus one: OR 0.40; 95% CI: 0.21, 0.79) or more than 2 rooms (OR 0.35; 95% CI: 0.20, 0.62; p trend=0.001). pfPR was significantly positively associated with keeping a dog inside or near the house (OR 2.01; 95% CI: 1.26, 3.21).

Table 5.

Multivariate associations of factors with asymptomatic Plasmodium falciparum malaria among healthy children 0-15 years old in northern Tanzania, surveyed between October 2015 and August 2016.

Characteristics	Odds ratio* (95%CI)	p†
Geographical
Region		0.18
Simiyu and Mwanza Regions	Ref
Mara Region	0.67 (0.35, 1.29)
Geita Region	1.34 (0.67, 2.70)
Parental characteristics
Mother’s education		0.01
Up to primary 4	Ref
Primary 5-7	0.55 (0.31, 0.96)
≥ Senior secondary school	0.10 (0.02, 0.55)
Father’s education		0.01
Up to Standard 4	Ref
Standard 5-7	0.85 (0.44, 1.64)
≥ Senior secondary school	0.14 (0.04, 0.49)
Mother’s occupation		0.94
Trader/Sales	Ref
Peasant farmer	0.86 (0.29, 2.59)
Manual laborer	1.16 (0.18, 7.50)
Father’s occupation		0.41
Trader/Sales	Ref
Peasant farmer	0.62 (0.23, 1.70)
Manual laborer	0.37 (0.08, 1.63)
Home characteristics
Distance of home from main road		0.02
Far from the main road	Ref
Near the main road	0.34 (0.15, 0.76)
In town or city	0.21 (0.03, 1.42)
Source of drinking water		0.38
Unprotected spring/well	Ref
Protected spring/well	0.73 (0.36, 1.47)
Public tap/piped household	0.61 (0.29, 1.31)
Connected to electricity grid		0.38
No	Ref
Yes	0.60 (0.19, 1.90)
Number of rooms in house		<0.01
1 room	Ref
2 rooms	0.40 (0.21, 0.79)
≥ 3 rooms	0.35 (0.20, 0.62)
History of fevers and hospital admission
Reported ≥1 fever due to malaria in the last 6 months		0.05
No	Ref
Yes	0.55 (0.29, 1.01)
Animals kept near or inside house
Chicken		0.68
No	Ref
Yes	1.15 (0.58, 2.29)
Dogs		<0.01
No	Ref
Yes	2.01 (1.26, 3.21)

CI = confidence interval

^*The ORs are mutually adjusted for all the variables found to be significant at p < 0.10 in the univariate analysis, including variables that are not significant in the multivariate model (p<0.05).

^†p for heterogeneity

Discussion

Our study found that nearly 40% of children in two regions in northern Tanzania had low-grade malaria antigenemia, consistent with previously reported high regional malaria transmission rates (1, 26) and eBL incidence (11, 13). The key findings are a high but heterogenous pfPR in northern Tanzania, significant inverse associations between pfPR and having parents with 5-7 or more years of formal education, residing in a household in more urbanized areas, and living in a house with at least two rooms. The results contribute important data to inform national goals for malaria control and provide important baseline data for studies seeking to precisely define malaria-related risk factors of eBL in northern Tanzania. The results presented here provide a useful basis for comparison to geographical and age-related patterns of eBL in our study area.

Although a previous study indicated a two-fold difference in eBL incidence in Mara versus Mwanza (11), we found comparatively less variation in pfPRs in these two regions. Furthermore, while eBL incidence across districts in these regions was quite heterogenous (11, 13), our pfPR results suggest less extreme geographical variation for malaria. However, direct comparison of our pfPR results and the previous eBL results should be done carefully because these studies of pfPR and eBL were not concurrent. The less extreme patterns of pfPR relative to those of eBL suggest that additional non-malaria-related factors may contribute to the marked variation in eBL incidence patterns. Understanding the nature and the role of those other co-factors, including Epstein-Barr virus (27), will be the focus of future studies in EMBLEM.

Our finding of a null association in the univariate analysis between pfPR and ownership of mosquito bed nets or using them the night before interview echoes our findings in Uganda (23) and Kenya (28) obtained using similar methodology as well as other reports in Tanzania (29, 30) and elsewhere (23, 28, 31). These results highlight an emerging concern that mosquito bed nets are losing their effectiveness as a key tool for malaria control, while the successful implementation of bed nets programs provides a false-sense of security. This trend may be due to non-use or improper use of bed nets (32, 33), deterioration with holes in nets (34–36), replacement of dominant mosquito vector that fed at night indoors with others that feed outdoors (37), or emergence of insecticide resistance (38–40). Further research is needed to understand the reasons underlying the null associations with mosquito bed nets, and to support the development of new strategies to expand and diversify insecticides deployed for malaria prevention (41).

Although regional variation in pfPR was apparent, no differences were noted in pfPR for children living near versus those living far away from the lakeshore. Similar results were obtained in Western Kenya when pfPR in the lake-endemic zone districts near versus those far away from the lakeshore was considered (28). Although these results contrast with those reported from studies conducted in swampy areas of Kampala, Uganda, which showed that malaria risk increases with proximity of a home to a swamp (42), both results are consistent with the notion that pfPR is influenced by the local ecology where mosquitoes breed, such as small pools of water and shrubbery, and not necessarily by large water bodies. Further research to characterize the local factors influencing malaria transmission is needed to guide larvicidal efforts to reduce local malaria transmission.

Our observation of strong inverse associations between pfPR and socioeconomic-related variables is consistent with the notion that broad-based government interventions to improve the economy of communities have a positive impact on malaria burden. Although we did not collect detailed information about the houses to help us definitively discuss the characteristics of large houses that are associated with lower pfPR, previous studies have shown protection with improved (metal) roofs, closed eaves, and screens on windows and doors decrease the number of mosquitoes that enter a house and, thereby reduce malaria transmission to household members(43–46). Our results suggest that rapidly urbanizing regions of Tanzania could improve their malaria control through attention to housing (47).

Our observation of a positive association between pfPR and keeping certain animals inside or near the house may be a chance finding. However, the positive associations reported above might also reflect changes in the predominant mosquito species and their associated behavior with respect to human biting patterns, e.g., biting outside the home rather than inside, that result in increased malaria transmission risk (48, 49). Further research is needed to evaluate the relationship between animals kept near or inside the house, mosquito behavior, and malaria transmission.

However, the results are cross-sectional, which limits inferences about the temporality of associations. Lack of contemporaneously information about mosquitoes circulating in the communities limits our ability to correlate epidemiological data with entomological patterns. We also acknowledge that the limit of detection (~200 parasites /uL) for RDT and malaria microscopy is not optimal for detecting all infections(50). More sensitive tests will be required to monitor the epidemiology of malaria as the burden of malaria in northern Tanzania nears levels where elimination becomes the goal (4). The strengths of our study include using a representative general population sample of children residing in northern Tanzania, having detailed risk factor information, and careful assessment of malaria prevalence. Although performing many statistical analyses is a weakness, our study was explorative, and the findings point to areas where further research is needed.

Conclusion

We observed that nearly 40% of children in northern Tanzania had malaria antigenemia. Antigenemia was inversely associated with socioeconomic-related factors, but not mosquito bed net ownership and use the night before survey. Our results are consistent with the notion that broad-based economic improvements enhance the impact of malaria control programs, and with the emerging concern that mosquito bed nets are losing their effectiveness in Tanzania.