Epidemiology of forest malaria in Central Vietnam: the hidden parasite reservoir

Study site and population

The study was carried out in four villages located in two communes (Tra Leng and Tra Don) situated in Nam Tra My district in Quang Nam Province (Central Vietnam) (Figure 1). Study villages were located in a remote forested valley accessible only on foot (five hours) or motorbike (two hours) on a mountain track from the nearest health centre in Tra Don commune. Villages were extremely scattered, with households grouped in clusters of four to 45 houses situated at variable distance from each other. The number of clusters varied by village with four clusters in Village 1, two in Village 2, nine in Village 3, and five in Village 4. All study clusters were served only by the CHC in Tra Leng since the one in Tra Don commune was too far. Village 3 and 1 were located along the way to and around the CHC, respectively, while Villages 2 and 4 were situated at 4- and 3 hours walking distance (for the farthest clusters) from the CHC. In addition, there was a river between the centre of the commune and Village 4 whose access was almost impossible during the heavy rains.

The population mainly belonged to the M’nong and Ca Dong ethnic groups living in very poor socio-economic conditions, mainly subsistence farming, practising slash-and-burn agriculture in forest fields with maize, manioc and rice. Malaria transmission is perennial with two peaks, one in May-June and the other in October-November, with the two main vectors species being Anopheles dirus sensu stricto and Anopheles minimus sensu stricto [10,11]. Malaria control activities are based on free-of-charge, early diagnosis and treatment with an artemisinin-based combination (ACT; dihydroartemisinin-piperaquine) and regular indoor residual spraying (IRS; alpha-cypermethrin) as bed net use was not very popular in the study area at the time of the survey (Nguyen Van Van, personal communication).

The Commune Health Centre (CHC) located in the centre of the commune (Village 1) was hardly accessible for Village 4 during the rainy season because of heavy rains and flooding. The local health staff (one midwife, three nurses, one microscopist, and one pharmacist) provided free-of-charge health care with the support of village health workers (VHWs).

Data collection

In February 2009 a full census of the study population (1,810 individuals) was carried out to collect household as well as individual socio-demographic data (gender, age, location, occupation, assets, distance to the fields, number of available bed nets per household, housing structure, etc.). Each resident in the study area was allocated a unique ID code. Each house was mapped using a geography position system device (Garmin eTrex Legend HCx Personal Navigator) [12].

In April 2009, the entire study population was screened for P.vivax infections to identify potential study participants for a cohort to be followed prospectively. This started by informing first all commune, village and household leaders on the objectives and study procedures and then the individual study subjects, who were all invited to be screened after oral informed consent. During the screening, each participant was interviewed for previous malaria symptoms during the previous 48 hours, the axillary temperature was checked and a blood slide collected for light microscopy (LM). Confirmed malaria infections were treated according to the national treatment guidelines.

In addition to blood slides taken during the screening, an additional blood sample was taken for haemoglobin measurement and for later molecular analysis (PCR) in a random sample of study participants (n = 327). This was done by randomly choosing one individual in each household after blindly drawing an ID number among those allocated to the house during the census. If the selected subject was temporarily absent, the survey team would return later; however, if the subject was absent for a long time or not willing to participate, a second drawing would be done. Survey participants (i.e., with additional blood samples) were asked to give their written informed consent (parent/guardian for children) after being explained the purpose of the additional sampling and investigations. Among these subjects, a face-to-face interview was done to collect data on the different outdoor activities in and outside the community, sleeping habits, as well as malaria prevention measures. For children under 12 years old, the parent/guardian would answer the questions.

Laboratory procedures

Thick and thin films were stained with a 3% Giemsa solution for 45 minutes. The number of asexual forms per 200 white blood cells (WBCs) was counted and parasite densities were computed assuming a mean WBCs count of 8,000/μl. Gametocytes were also counted. A slide was declared negative when no parasite was seen after counting 1,000 WBCs. All slides were read independently by two expert microscopists. In case of discrepant results, they re-examined the slide together until agreement was reached. Quality control of blood slides was done on all positives and 10% of negative blood slides by a senior laboratory technician at the National Institute of Malariology, Parasitology and Entomology (NIMPE), Hanoi; in case of disagreement, a second senior technician would re-read the slide until an agreement was reached.

Haemoglobin concentration was measured with the HemoCue Hb 301 device following the manufacturer’s instructions [13]. Filter paper blood samples (FPBS) were dried outside in direct sunlight and kept in individual, sealed, plastic bags containing silica gel. All FPBS were stored at 4°C in the CHC refrigerator before being shipped to NIMPE, Hanoi, where they were kept at −20°C. DNA extraction was done using the QIAamp DNA Micro Kit (Qiagen, Hilden Germany), and a species-specific, semi-nested, multiplex PCR (SnM-PCR) was performed to detect P. falciparum, P. vivax, P. malariae, and P. ovale [8]. The PCR products (5 μl) were subjected to electrophoresis on a 2% agarose gel in 0.5X TAE buffer for 90 minutes at 100 V. The gels were stained with ethidium bromide and visualized with ultraviolet light. The sizes of the PCR products were compared with a standard 100-basepair DNA ladder (Fermentas, Burlington, Ontario, Canada) and positive controls of each Plasmodium species. Cross-contamination during handling was checked for by implementing negative controls in each step from extraction to the nested PCR step. Quality control was done on 10% of the samples for which the SnM-PCR was repeated blindly by a senior technician. In case of discrepancy, the sample was re-analysed until agreement was reached.

All survey samples were analysed for G6PD deficiency at the Shoklo Malaria Research Unit, by genotyping for the Viangchan mutation following a modified protocol published by Nuchprayoon et al. [14]. DNA was extracted using the Saponin-Chelex method [15]. Genotyping for the Viangchan mutation (871G > A) was performed by PCR/RFLP method using published primers [14] and MyTAq™ DNA polymerase (Bioline, UK) with the following amplification conditions: initial denaturating step at 95°C (5 min) followed by 30 cycles of 95°C (30 sec), 57°C (20 sec), and 72°C (15 sec) and final elongation step at 72°C for 7 min. Amplified fragments were digested with XbaI enzyme and visualized on a 3% agarose-nusieve gel. Quality control was performed on 10% of randomly selected survey samples; in case of disagreement, the sample was re-analysed by another senior technician.

Data management and statistical analysis

Sample size: According to the provincial malaria station data on surveys carried out in April-May, the overall parasite rate was around 16% (ranging from 5 to 39% across hamlets) and the prevalence of P. vivax at 9%. The sample size was calculated by assuming a minimal prevalence of 9%, with 3% precision at 5% significance level and adding 10% security margin; a total of 330 individuals were needed for the survey (“CSample” command/EpiInfo6). Therefore, to simplify sampling procedures, one individual in each house visited during screening was randomly selected to be included in the survey.

Data were double-entered and cleaned using Epidata version 3.1 free software [16]. The data set was analysed using STATA version 11 (Stata Corp, College Station, TX, USA). Descriptive statistics were used to compute baseline socio-demographic characteristics as well as malariometric indices by village, and significant differences were tested for using either a Chi-square test or Student t-test as required, and a p-value <0.05 was used as cut-off for significance.

Three different variables for livestock ownership (number of i) buffaloes, ii) cows, and iii) pigs) were considered as the best proxy (after discussion with household leaders) for the economic status of the households as all inhabitants were subsistence farmers and generally poor. In order to aggregate multiple variables to a single measure of economic status, a principal component analysis was performed [17]. Using the factor scores from the first principal component as weights, an index was created for the economic status of each household then the index were categorized by dividing the score into tertile.

The survey design (survey dataset) was taken into account using the svy- command in STATA, with villages as strata, and household sizes as p-weights. A survey logistic regression (“svy” command in STATA) was used to carry out a multivariate adjusted analysis for the risk of malaria infection (determined by PCR). Moreover, a classification tree analysis (CART; Salford Systems Inc, CA, USA) was performed to explore the relationship and rank the relative importance of risk factors for all malaria infections identified by PCR, as well as for patent infections only (detected both by PCR and microscopy). Sub-patent (or sub-microscopic) malaria infections are detected by PCR only. The CART analysis is a non-parametric method enabling more direct and flexible analyses since, unlike logistic regression models, it allows for co-linearity and multiple interactions between different independent variables [18]. Briefly, the building of the classification tree starts with the root node, which contains the entire set of observations. CART then finds the best possible variable to split the root node into two child nodes, by identifying the best splitting variable that maximizes the average ‘purity’ (homogeneity) of the two child nodes. To improve the stability of the CART model, a ten-fold cross-validation method was applied, and the best tree was selected by choosing the smallest tree within one standard error of the minimum error. CART also provides a ranking power of each predictor variable.

Ethical clearance

Ethical clearance was obtained from both the ethical committee of NIMPE in Hanoi and the University of Antwerp. The fundamental principles of ethics in research on human participants were upheld throughout the project. All study participants gave their informed consent after being explained the study procedures as well as their right to withdraw without prejudice for themselves or their families.