Malaria epidemiology in the Ahafo area of Ghana

The study was carried out in four districts in the Brong Ahafo Region of Ghana; the Asutifi District (between latitudes 6°40' and 7°15' North and Longitudes 2°15' and 2°45' West), the Tano North and South (Tano N/S) Districts (between latitudes 7°00'N and 7°25' N and between longitudes 1°45 W and 2°30 W), and the Techiman municipality (between latitudes 7°31'N and 8°0' N and between longitudes 1°60 W and 2°00 W). The study areas are in the wet semi-equatorial forest zone where the mean annual rainfall is about 1,200 mm per annum. Farming of cash crops, such as Cocoa, Oil Palm, Coffee are the major economic activity in the study communities. All health facilities in the study area implement both clinical and public health services under the supervision of the Brong Ahafo Regional Health Directorate. Malaria is the leading cause of morbidity and mortality in the region. About eighty percent of residents in the study area belong to the Akan ethnicity.

A household cross-sectional study was carried out between July 2006 and March 2007 to determine baseline household morbidity and malaria transmission prior to the start of mining activities in the area. The districts in the mining area were categorized into two; the Impact area involving communities within the Asutifi, and Tano N/S districts whose health and/or socio-economic indices are likely to be affected by mining activities being undertaken and located within approximately 25 Km radius from the mine sites. The Non-impact area representing communities within the Asutifi, Tano North or Tano South districts that are not likely to be affected by the mining activities and located more than 25 Km from the mines site (Figure 1). The Techiman municipality was purposively selected as a control area without mining activities but under the same local political and health administration as the other districts. The control area was selected for future comparisons after the beginning of mining activities. Within each area, communities were purposively selected based on geographical location to ensure representation of all locations within each area. Eleven (11) out of twenty (20); thirteen (13) out of nineteen (19) and eight (8) out of eighteen (18) communities were selected to represent the study areas in the Asutifi, Tano N/S and Techiman areas respectively.

All households in the selected communities were enumerated and encoded into a database. Each household member in the community including children less than five (5) years was given a unique identification number as part of the household morbidity survey. All households who were respondents for the household morbidity survey and with children less than five (5) years in the database formed the sampling frame. One child less than five (5) years was randomly selected per household using a computer software for anthropometric assessment, hematological and malaria parasitaemia.

Household ownership of ITNs and other household characteristics were assessed using a pre-tested structured questionnaire which were administered by trained fieldworkers. The quality of the ITNs was assessed by direct observation for holes.

Malaria parasitaemia was determined by collecting finger prick blood samples from participants to prepare thick and thin blood smears. The smears were stained with 10% Giemsa and air-dried. They were examined under the microscope using ×10 eyepieces and ×100 objective. Malaria parasites were counted against 200 white blood slides on the thick smear. A slide was declared negative only after 200 oil immersion fields of the thick smear had been examined. Parasites were expressed per microlitre, assuming a total white cell count of 8,000 per microlitre. Species identification was done using the thin smear. A sub-sample (10%) of the blood slides were randomly selected and sent to an independent expert microscopists for examination and the results compared with > 90% concordance.

Part of the blood from the finger prick was collected into Hemocue microcuvettes and haemoglobin levels were determined using a calibrated hemocue (HemoCue GmbH, Germany). All children found to be unwell, parasitaemic or anaemia during the survey were referred to a study clinician and treated.

Entomological surveys were carried out in the Asutifi and the Tano N/S districts from November 2006 to August 2007. The dry season period (Nov. 2006 - Feb. 2007) and wet season period (May 2007 - Aug. 2007) were purposefully planned to compare transmission with or without the influence of rainfall in a mining area. No entomology work was carried out in the Techiman district as similar studies were carried out in the Kintampo district, an adjacent district with similar rainfall pattern in 2004 [3]. Centre for Disease Control (CDC) light trap catches were performed concurrently on each night of mosquito collection in both impact and non-impact communities. Index persons were randomly selected from the Asutifi/Tano household enumeration database. A total of about 1,100 possible compounds were randomly selected from enumerated households in the Asutifi and Tano N/S districts to receive mosquito CDC light traps. The house of the "index" person was the house of choice to set traps. Traps were hung at the foot end adjacent to the "index" net of the "index" person. Untreated mosquito nets were provided to household members whose rooms were used on the night of trap setting to ensure they were protected while ensuring that mosquitos were not repelled. This method was approved by ethics committees. Four traps were set weekly and mosquitoes collected were chloroformed and morphologically identified using Anopheline morphological identification keys [10]. They were stored in 1.5 ml eppendorf tubes and transported on weekly basis to the Kintampo Health Research Centre laboratories for ELISA as per the procedures of Wirtz et al[11]. An average cut-off point of 0.2 nm absorbance wavelength was considered positive. Positive samples were re-tested and PCR analysis performed as described [11‐13]. The entomologic inoculation rate (EIR) was calculated as the product of the proportion of Anopheles positive by ELISA also termed as sporozoite rate (SR) and the Human Biting Rate which is estimated as the geometric mean of Anopheles caught per CDC light traps set.

A total of 200 morphologically identified Anopheles gambiae s.l. specimens were further analysed by polymerase chain reaction (PCR) in order to establish members of the species complex and molecular forms composition [14].

All data collected in the field or the laboratory were logged for traceability, and then batched for double data entry and processing using Microsoft^® Visual FoxPro 6.0. All data management processes were done at the computer laboratory of the Kintampo Health Research Centre. Data analysis was done using StataCorp Stata 10, TX USA. Averages and proportions were used to provide descriptions of age- and parasite prevalence, density, infection rates and the proportions of parasite clones that persisted during the survey. Socio-economic status was calculated for each household using World Bank asset scores for Ghana, based on ownership of a television, radio, refrigerator, car, motorcycle, bicycle, farm, electricity in house, ceiling, floor cover, and number of individuals/sleeping room. The anthropometric indices height-for-age (HA), weight-for-age (WA), weight-for-height (WH), were expressed as Z-scores using the WHO Anthro for personal computers, Version 3.1, 2010: Software for assessing growth and development of the world's children. A child was identified as being malnourished if s/he scored < -2 in one of the HA, WA, or WH indices.

Malaria parasitaemia, anaemia (haemoglobin < 10 g/dL) and ITN usage were the main outcome variables. Odds ratios were calculated with their 95% confidence interval to assess the relationships between exposure and outcome variables using univariate and multivariate logistic regression models. The Wald's test and likelihood ratio test were used to estimate P values for exposure variables with two (2) levels and more than two (2) levels respectively in both the univariate and multivariate logistic regression models. P values less than 0.05 were considered statistically significant. Missing values were not included in the logistic regression models.

Ethical approval to conduct the study was granted by the Kintampo Health Research Centre Institutional Ethics Committee. Community entry involved explaining the study to key community opinion leaders followed by community durbars/meetings. At these meetings, the study's aims, objectives, risk and benefits were explained to all participants. A signed/thumb printed written consent was obtained from each respondent. All data collected are kept confidential.

A total of 1,671 households with a child less than five years were selected. Response rate was approximately 100%. About 50% of the household heads were males (Table 1).

Majority of household heads were between ages 30-39 years (27.8%) and 40-49 years (27.3). Majority of children (63.6%; 95% CI 61.2 - 65.9) in the households were ≥ 24 months. Health insurance coverage was 42.3%, (95% CI 40.5 - 45.3).

A total of 1,671 children less than five years were surveyed for malaria parasitaemia and anaemia. The prevalence of any malaria parasitaemia was 22.8% (95% CI 20.8 - 24.9). Plasmodium falciparum constituted 98.1% (95% CI 96.2 - 99.2) of parasitaemia. Plasmodium malariae was uncommon -1.9% (95% CI 0.81 - 3.8); no Plasmodium ovale or Plasmodium vivax was identified; there were no mixed infections. The geometric mean P. falciparum asexual parasite count was 1602 (95% CI 1,140 - 2,252) and 1,195 (95% CI 985 - 1,449) among children < 24 months and ≥ 24 months respectively. The proportion of children with parasite counts at various cut-off points was similar among the two age groups (Figure 2). Majority of the children (< 24 months: 74%, 95% CI 64.4 - 82.9; ≥ 24 months: 82.9%, 95% CI 76.7 - 86.1) had low parasite counts < 5,000/μL. A univariate logistic regression model analysis was fit prior to multivariate regression models. Malaria parasitaemia was significantly associated with the socioeconomic quintile (p < 0.001) with the strength of the association decreasing along the better socioeconomic quintiles. The least poor were significantly less likely to have malaria parasitaemia compared to the poor, OR 0.40 (95% CI 0.27 - 0.60) (Table 2). Health insurance membership protected against malaria parasitaemia (OR 0.51, 95% CI 0.40 - 0.66, p < 0.001). Household ITN ownership was weakly associated with malaria parasitaemia (OR 0.80, 95% CI 0.62 - 1.03, p = 0.08). Children ≥ 24 months were significantly more likely to have parasitaemia compared to children < 24 months (OR 1.95, 95% CI 1.50 - 2.53, p < 0.001) (Table 3). In the multivariate logistic regression model, there were strong evidence of associations between malaria parasitaemia and household membership with health insurance (p = 0.001); child age (p < 0.001) and anaemia (p < 0.001).

About 68.7% (95% CI 66.3 - 70.9) of households own at least one ITN. The median ITN per household was 2 (range 1 - 7). About 18.1% (95% CI 15.9 - 20.5) of ITNs observed for quality had holes in them. About 86.4% (95% CI 84.3 - 88.4) of household that owned ITNs had at least one person sleeping under an ITN the night before the interview. Majority of ITNs (64.9%, 95% CI 62.1 - 67.7) were obtained free of charge as part of a national campaign while 26.4% (95% CI 23.8 - 29.0) bought their ITNs on the open market at commercial prices. There was no significant association between household ITN ownership and household head characteristics such as sex and educational level in the univariate logistic regression model (Table 6). There was however a significant association between ITN ownership and attributes such as health insurance membership and area of residence that remained significant in the multivariate logistic regression models. Households with children ≥ 24 months were significantly less likely to own an ITN (Table 7)

A total of 5,257 mosquitoes were caught from 818 light traps set; Anopheles gambiae comprised of 64.1% (n = 3,371) of the mosquitoes caught, Anopheles funestus 4.3% (n = 227) and non-Anopheles species 31.6% (n = 1,659) which were predominantly Culex and Aedes species of mosquitoes. A few Mansonia species were recorded in Asutifi (4) and Tano N/S districts (2).

EIRs for both Anopheles gambiae and Anopheles funestus were generally high. In the impact and non-impact areas of the Asutifi district, the median EIR was 46 ib/p/m (range 0-826 ib/p/m) and 51 ib/p/m (0 - 216) in the impact and non impact areas respectively. The highest was recorded in June 2007 at both impact and non impact areas respectively (Figure 4). EIRs in Tano N/S districts were also high, median 65 ib/p/m (range 48 - 219 ib/p/m) and 102 (range 14 - 669 ib/p/m) in the impact and non-impact area respectively with the highest in May 2007 (Figure 5).

The EIR of Anopheles gambiae were generally higher than Anopheles funestus. In Tano N/S alone, the median EIR for Anopheles gambiae alone was 73 ib/p/m (range: 39 - 438 ib/p/m) compared to Anopheles funestus median 0 ib/p/m (range 0 - 5 ib/p/m) alone (Figure 6).

Mosquitoes randomly selected and amplified by PCR were all An. gambiae sensu-stricto as confirmed by the size of the amplified DNA fragment of 390 base pairs. One hundred out of the 200 mosquitoes identified as An. gambiae s.s. were randomly selected for the identification to the molecular form by enzyme digestion with Hha 1. All 100 mosquitoes were S-forms shown by restriction analysis.

Two hundred mosquitoes were randomly selected to determine knockdown resistance (kdr) mutations. One hundred and ninety three out of the 200 selected mosquitoes (97%) possessed the kdr mutation (kdr+), whilst seven (3%) did not (kdr-).