Study area

The study was conducted in three villages in Tanga region, North-Eastern Tanzania. The three villages were Mgome (5°12'S, 38'51'E) at an altitude of approximately 200 meters, Ubiri (4°72'S, 38°29'E) at an altitude of approximately 1,200 meters, and Magamba (4°75'S, 38°29'E) at an altitude of approximately 1,700 meters.

The climate in the area is characterized by variations in rainfall and temperature related both to season and altitude [12]. The long rainy period occurs during April-May, while short rains occur in November-December. Mean daily temperatures are highest in January and lowest in July. Generally, the malaria transmission season peaks just after the rainy seasons with most consistent transmission in lowland sites from April to July. Previous studies have reported parasite prevalence rates to be in the ranges of 79–90% in the lowlands, 27–46% at intermediate altitudes and 8–16% in the highlands [10]. Entomological surveys in the study areas have shown that Anopheles gambiae is the most prevalent vector in the lowlands, while Anopheles funestus predominates in the highlands [10]. The entomological inoculation rates (EIR) have been reported to be in the range between 91–405 in the lowlands, and between 1.8–34 at intermediate altitudes [10]. In the highlands, mosquito densities are too low to allow reliable EIR measurements, but an EIR of 0.03 has been extrapolated [10]. Villagers living at low and intermediate altitudes perceive malaria as a major problem among both children and adults, but at the highest altitudes villagers consider that malaria is not a major part of the disease burden in either adults or children. There is little difference in treatment seeking behaviour for febrile illness between the altitudes. Treatment is generally sought for symptoms rather than for the disease and first treatment is almost universally an anti-pyretic drug bought from local shops (Caroline Jones, unpublished data). For all three villages, the nearest health facility is located within a distance of 13 km. Mgome is served by Umba Dispensary (10 km), Masaika Dispensary (5 km), Mkuzi Health Centre (7 km) and Muheza Designated District Hospital (14 km). Ubiri village is served by Lushoto District Hospital at a distance of approximately 13 km. Magamba village has a government and a private missionary dispensary both within the village, and is served also by Lushoto District Hospital at a distance of about 15 km. At the time of the study, sulphadoxine-pyrimethamine (SP) was the first-line treatment for uncomplicated malaria in Tanzania. It has been documented that the level of SP resistance is high in the Mgome area [13], whereas the situation has not been monitored previously in Ubiri and Magamba.

Land use in the lowland areas is characterized by subsistence farming of maize, rice, bananas, beans, cassava, coconuts, fruits and other crops, as well as large-scale production of sisal. In the highlands, there is subsistence farming, mainly of maize, beans, bananas, potatoes, cabbages, tomatoes and fruits, and also large-scale production of tea and coffee.

Study population

Prior to the study, census surveys were done in each village and study individuals randomly selected from a census list. Mgome village is inhabited mainly by the Bondei tribe (60%), while Ubiri and Magamba are inhabited by Sambaa at 97% and 57%, respectively. The aim was to recruit a total of 250 individuals below the age of twenty years from each village, distributed in different age groups as follows: 0–1 year: n = 25, 1 year: n = 25, 2 years: n = 25 3 years: n = 25, 4 years: n = 25, 5–6 years: n = 25, 7–9 years: n = 25, 10–14 years: n = 40 and 15–19 years: n = 40.

Cross-sectional surveys

Malariometric surveys were conducted in each village in April, July and September 2001. During the first survey, the purpose of the study was explained and consent to participate obtained from each study individual or their parents/guardians. Baseline demographic data were collected together with a history of migration and recent movements. The use of malaria preventive measures was also recorded. A history of recent illness was obtained, emphasizing symptoms suggestive of malaria. Physical examination on signs related to malaria such as temperature, pulse, spleen size, pallor and respiratory rate was conducted. Axillary temperature was measured using digital thermometers. Height, weight and upper-arm-circumference were recorded for estimation of nutritional status. For any individual diagnosed with mild disease, appropriate drugs were administered in the field. Individuals with symptoms of malaria were treated with SP. Participants with severe disease were referred to the nearby hospital.

Five millilitres of venous blood were collected from study individuals aged three years and above into vacutainer tubes containing citrate buffer. For children below three years, 300–400 μl of capillary blood from a fingerprick were collected into eppendorf tubes containing EDTA. The haemoglobin (Hb) of each participant was measured from drops of blood using a HemoCue^® photometer (Ångelholm, Sweden). Whole blood was used to prepare thick and thin blood smears for malarial microscopy. These were stained with 10% Giemsa stain for 15–20 minutes after fixing thin smears with methanol. Asexual and sexual parasites were counted against 200 and 500 white blood cells, respectively. The differentiation of malaria parasite species was confirmed by microscopy of thin smears. A blood smear was declared negative only after examination of 200 high power fields. The density of asexual parasites was calculated assuming 8000 leucocytes per μl of blood and expressed as parasites per μl.

During the first cross-sectional survey, study participants were asked to collect stool and urine specimens in special containers. Direct smear-technique was used to check for the presence of hookworm ova and other intestinal parasites. A pinhead of stool was collected, put on a slide and emulsified in a drop of normal saline. A cover slip was then applied and the slide examined using low-power microscopy.

Longitudinal monitoring of febrile episodes

Local village helpers (two community members per village) and health workers at nearby health facilities performed passive case detection during the 6-month study period. The village helpers were provided with first-line antimalarial drugs (SP), paracetamol, slides, blood lancets, treatment charts, febrile case detection forms and storage boxes. Villagers could seek treatment at any time from these helpers. Patients with symptoms of malaria were treated with first-line antimalarial drugs or, if they had severe symptoms or did not respond adequately to the first-line treatment, they were referred to a health facility. Prior to treatment the village helpers collected clinical information and a malaria blood smear.

At each nearby health facility, two permanent staff members monitored study participants seeking medical treatment at the facility. If a study participant presented at the facility with history of fever and/or an axillary temperature ≥ 37.5°C, a form was completed and a blood smear collected. Once per month active febrile case detection was undertaken by the research team. During active case detection, each study participant was seen by a trained physician and a blood smear was taken from any study participant reporting a history of fever within two days and/or those who had an axillary temperature ≥ 37.5°C

Case definitions

Anaemia was defined as haemoglobin < 11.0 g/dl [14, 15]. To adjust for the physiological effect of altitude on haemoglobin concentration, a correction factor was calculted with haemoglobin values being normalized to sea level for direct comparison between the study villages. The correction factor assumed a linear relationship between increasing altitude and haemoglobin, although the relationship may not necessarily always be exact [16]. For Mgome (200 m), the correcting factor was a reduction of 0.1 g/dl, for Ubiri (1,200 m) the factor was 0.8 g/dl and for Magamba (1,700 m) the factor was a reduction of 1.0 g/dl. Febrile malaria episodes were defined as an axillary temperature ≥ 37.5°C and /or a history of fever within the previous 48 hours in the presence of asexual P. falciparum parasites above a defined density cut-off level. Many individuals carried low density asymptomatic parasitaemia, and fever among parasitaemic individuals may also have been caused by other illness [17]. Thus, to account for the variation in levels and point prevalence of asymptomatic parasitaemia between study villages [18–20], as well as the different age groups involved in the study [21], different P. falciparum density cut off levels were applied in each village. To balance between sensitivity and specificity in diagnosing a febrile malaria episode, we aimed at a febrile malaria case specificity >80%. In Magamba (the low transmission village), a cut-off of 40 parasites/μl was applied, while cut-offs of 1000 parasites/μl and 5000 parasites/μl were used in Ubiri (the moderate transmission village), and Mgome (the high transmission village), respectively. Age-specific incidence rates of febrile malaria episodes were calculated as the number of episodes divided by the number of days that individuals in the age group were at risk during the follow-up. After a febrile malaria episode an individual was censored for 28 days [6]. The effect of using different parasite density cut-offs in the definition of a febrile episode was evaluated by not applying a cut-off in the definition or by applying age specific cut-offs [21].

Statistical methods

All data were double-entered into a database in Epi-lnfo Version 6.04d (CDC, Atlanta, USA) and statistical analyses were performed with Stata version 8 (Stata Corporation, Texas, USA). Univariate analyses and multivariate logistic regression were performed to determine risk factors for anaemia and febrile malaria episodes.

For Mgome village, a logistic regression model was developed to determine whether the result of the first slide reading in April could be used to predict the subsequent risk of developing anaemia or febrile malaria during the following six months of morbidity surveillance. In this model, P. falciparum parasitaemia was categorised as no parasitaemia if no parasites were detected microscopically, low-density if parasitaemia was between 40 parasites/μl and 4999 parasites/μl, and high-density if the level was above or equal to 5000 parasites/μl. Thus, the first slide reading of individuals who did not have fever/had normal haemoglobin levels at enrolment was used to predict the risk of developing a subsequent episode of malarial fever/anaemia.

Ethical considerations

Ethical clearance was granted by the Medical Research Co-ordinating Committee of the National Institute for Medical Research, Tanzania. Prior to the study, meetings were held with local authorities and with the villagers in each study village, during which the aims of the study were explained. Informed consent documents for the study were prepared in English and translated into Kiswahili before administration to both village leaders and participants. Written informed consent to participate was obtained from each study individual or from his/her parents or guardians. Study individuals were free to withdraw from the study at any time without giving any reasons, or being disqualified from any medical services that were provided to all villagers throughout the study period. At the end of the study, preliminary findings were presented at village meetings.

Prevalence and densities of Plasmodium species and other infections

As expected, P. falciparum prevalence and parasite densities (Figure 1) were higher in Mgome than in Ubiri and Magamba (trend test, z = 15.64, p < 0.001). In Mgome, the carrier rate was particularly high for children aged 1–9 years and then declined in the older age groups (trend test, z = -3.2, p < 0.001). In Ubiri, carrier rates were low in infants, peaked at the age of two years, but showed little variation in the age groups between 4 and 19 (Figure 1). Although carrier rates in Ubiri were slightly higher in April than in July and September, there were no marked seasonal changes in carrier rate by age in any of the villages. The parasite densities in those carrying parasites did not differ between villages after the age of six years. Among the under fives, children from Mgome and Ubiri carried higher levels of parasitaemia in July than in April and September 2001 surveys. Interestingly, between April and July surveys, there was a marked difference in the age-specific pattern of parasite density in Mgome. In April, the peak parasite density was noted in age group of 2 years, whereas the youngest had the highest parasite density in July surveys. Based on these findings, we categorised Mgome as a high transmission (holoendemic) village, Ubiri as a moderate transmission (mesoendemic) village, and Magamba as low transmission (hypoendemic) village. P. falciparum was the most predominant malarial parasite accounting for more than 95% of all malaria infections. The other malaria species were mainly found as mixed infections. The April prevalence rates of Plasmodium malariae in Mgome and Ubiri were 8.3 % and 3.9%, respectively, while these rates for Plasmodium ovale was 1.0% and 0%. In Magamba, only P. falciparum was found.

A total of 492 individuals from the three villages submitted stool and urine samples, which were investigated for worms. Worms were found in 35.0% of study participants living in Mgome, and in 29.2% and 7.8% of individuals from Ubiri and Magamba, respectively.

In Mgome, spleen enlargement was common (about 49.21%) and associated with age (Spearman rho = 0.238, p < 0.001) while in the two other villages the prevalence of splenomegaly was low and with no distinct age-pattern (data not shown). This distribution of splenomegaly remained stable during the study period.

Haemoglobin levels and anaemia

Haemoglobin levels were measured in all 759 individuals during enrolment and among those who reported for the subsequent cross-sectional surveys in July and September 2001. Regardless of the season, haemoglobin levels increased with both altitude and age (Figure 2). Univariate analysis indicated that age, altitude of residence, and presence of P. falciparum parasitaemia were associated with anaemia (Table 2) and this was supported by multivariate analyses in which P. falciparum was the only parasitic infection showing a statistically significant association to anaemia (Table 2).

Malaria morbidity during follow-up

Of the 759 individuals enrolled in the study, 669 (88%) adhered to the follow-up scheme and were included in the analysis of febrile episodes. Loss to follow-up was due to death (three individuals) or emigration. Using the village-specific density cut-off described above, 54 individuals had febrile malaria episodes in Mgome, 10 in Ubiri and none in Magamba during the six-month follow-up period. The mean age of febrile malaria individuals was 1.97 years (95% Cl: 1.50, 2.59) and 3.23 years (95% Cl: 1.59, 5.86) for Mgome and Ubiri, respectively. Children below five years carried the major burden of febrile malaria episodes in Mgome (Figure 3, panel A). The data was also analysed using age-specific parasite cutoffs [21] in the case definition (data not shown) and using a definition in which all fevers accompanied by a positive slide were considered a febrile malaria episode (Figure 3, panel B). The latter definition increased the incidence rates in Mgome and Ubiri slightly, but the overall conclusion that the incidence rates were by far the highest in the children under five years living in Mgome was not affected by the case definition used.