Malaria remains a major public health problem and a disease of poverty. In areas of stable and relatively high transmission, pregnant women and their newborn babies bear a high burden of malaria morbidity and mortality [1].

In Zambia, sulphadoxine-pyrimethamine (SP) is used for intermittent preventive treatment in pregnancy (SP-IPTp) during the second and third trimester. Until recently, malaria in pregnancy was treated with quinine, which has now been replaced by artemether-lumefantrine (AL) for women in the 2nd and 3rd trimester in pregnancy; quinine is used for the management of malaria during the 1st trimester and for severe cases [2, 3]. To provide additional evidence on safety and efficacy of artemisinin-based combination therapy (ACT) during pregnancy, a multicentre trial (PREGACT, Clinical Trials identifier NCT00852423) is currently on-going in four sub-Saharan African countries, including Zambia.

The Zambian study site is located in Nchelenge district, Luapula province, at the border with the Democratic Republic of Congo (DRC). This is an area of intense malaria transmission as the prevalence of malaria infection in children aged <5 years was 21.8% in 2008 (national average 10.3%) and that of anaemia (Haemoglobin [Hb] <11 g/dl) 55.9% (national average 49.0%) [4]. In 2010, the prevalence in under 5 years was as high as 50.5% in some age groups and anaemia at 79.7% [5]. The same malaria indicator survey provided information on the coverage of key malaria interventions, i.e. prompt effective case management, insecticide-treated mosquito nets (ITNs), indoor residual spraying (IRS), and IPTp [6]. Considering that some of the end points of the trial on ACT in pregnancy, e.g. incidence of recurrent infections, can be influenced by the local malaria endemicity [1] and that no new information had been collected since 2008, a malariometric survey was carried out in Nchelenge district in March-April 2012 [7, 8].

The rationale of this study was that it gives a more recent data on the prevalence of malaria in a high endemic region. In places like Zambia, where overall figures indicate that malaria is declining, there are still pockets of high endemicity, which are complicated by perfect breeding grounds for anopheles mosquito and cross-border human movements. This could affect national and global malaria control strategies. The primary objective was to estimate the prevalence and risk factors for malaria in Nchelenge, Luapula Province, Zambia.

Nchelenge district, Luapula province, is located on the swampy shores of Lake Mweru, and has a population of 178,000 inhabitants, mostly peasant farmers and/or fishermen. The district has one first level hospital, ten rural health centres and two health posts. The survey was done in the catchment areas (total population of 43,105) of the two health centres, Kashikishi and Nchelenge, where the PREGACT trial is currently implemented.

The household was taken as the sampling unit. The households in Nchelenge and Kashikishi health centre catchment areas were previously numbered using the Global Positioning System (GPS). Then the households were selected according to a pre-defined, computer-generated list of random numbers [9]. In selected households, all children below the age of 10 years were included in the survey after having obtained the signed informed consent from the parent/guardian. Two questionnaires were administered, the household questionnaire and the parent/guardian’s questionnaire which were adapted from those recommended for the malaria indicators survey [10, 11]. The household questionnaires include information on basic demographic and socioeconomic characteristics of the households.

The sample size was determined by assuming that malaria prevalence was 22% [4]; using Roasoft sample size calculator [12], 261 sampling units or households were needed to estimate such prevalence at 5% precision with a 95% confidence interval. Nevertheless, because of possible clustering within households, was assumed a conservative design effect of 2 was assumed, and a 20% adjustment was made for non-response (from household refusals or abandoned households). Based on these assumptions, it was estimated that 630 households would provide an estimate of prevalence with the desired precision.

All children <10 years of age in the selected households were asked to be included in the survey. Each study, subject was assigned a unique identifier for identification of biological samples and questionnaire data and to guarantee confidentiality. Information on bed net use, which was by observation, including long-lasting insecticidal nets (LLIN) or previous indoor residual spraying (IRS) was collected onto the structured questionnaires. A blood sample for a malaria rapid diagnostic test (RDT), (SD BIOLINE, Malaria Ag P.f., Standard Diagnostic Inc, Korea), thick blood film and for measuring haemoglobin (Hb) (Hemocue 301 machine, Angelholm, Sweden) was collected by finger prick. Children with fever (axillary temperature ≥37.5°C) and a positive RDT or only a positive RDT were treated with AL, the first-line treatment in Zambia. Children found to be sick or anaemic were referred to the health care centres for further management.

Thick blood films were stained with 10% Giemsa for 10 minutes. Two hundred high power fields were read before declaring the slide negative. The parasite density was determined by counting the number of parasites against 200 white blood cells (WBC) and assuming 8,000 WBC per μl. All stained slides were read by two independent microscopists unaware of the RDT results. In case of discrepant results between the two independent microscopists, a third microscopist read the slide until reaching a satisfactory agreement on positivity/negativity or the parasite density.

Data were double entered, cleaned and analysed using Epi Info™ (Center for Disease Control and Prevention, version 7.1.1.14). Further multivariable logistic regression analyses were performed using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA). Mean and standard deviation were computed to describe the distribution of continuous variables, while the distribution of categorical variables was described using counts and proportions. Geometric mean and median were computed for continuous variables that did not follow a normal distribution: parasite densities and Hb concentrations. Continuous variables were compared between age groups using t-tests and categorical variable proportions by χ² test or Fisher’s exact test. To account for uncertainty in the prevalence estimates, the 95% confidence intervals were computed. Anaemia was defined as Hb < 11 g/dl and created a dichotomous variable where Hb ≥ 11 g/dl was defined as non-anaemic. The risk of anaemia and malaria was estimated by fitting a multivariable logistic regression model and adjusted for possible risk factors to compute odds ratios (OR) and 95% confidence intervals (95% CIs). The variables included in the multivariable logistic regression model were selected a priori via extensive search of peer-reviewed literature. A two-sided p-value of ≤0.05 was used as threshold for statistical significance. The ‘survey logistic’ procedure in SAS 9.2 (SAS Institute Inc., Cary, NC, USA) with the option ‘clustering’ to account for the design effect of clustering within households.

Before the survey, the study protocol was approved by the scientific and technical committee (STC) of Tropical Diseases Research Centre (TDRC), the TDRC Ethical committee and the Ministry of Health. Participation in the survey was voluntary and participants were free to withdraw any time during the capture.

669 households were screened, 39 households were not responsive (21 were not available while 18 refused to participate). There was a 630/669 (94.2%) response. At this rate, the results presented are less biased. Out of the households that were responsive majority reported tubed well or borehole as main source of drinking water (87.5, 95% CI, 84.6-89.9). 90.0% (95% CI, 87.3-92.2) households had open pits or pit latrines without slabs as toilet facilities. 96.0% (95% CI, 94.1-97.4) households reported charcoal as the main source of household fuel. Most houses were made of finished brick walls (84.8%, 95% CI, 81.7-87.4) and grass thatched roofs or sticks and mud roofs (94.6%, 95% CI). In terms of owning durable goods, 83.5% (95% CI, 79.7-87.6) households owned a landline telephone or a cell phone, while a majority (64.3%, 95% CI, 60.4-68.0) also owned a bicycle (Table 1).

The prevalence of malaria infection was 30.2% (236/782), the large majority (98.0%, 231/236) Plasmodium falciparum and the remaining ones Plasmodium malariae (Table 2). No mixed infection was found. Plasmodium falciparum prevalence was the lowest in the <1 year old children while it varied between 22.7% (1 years old) and 31.8% (7–10 years old) in the other age groups (Table 2 and Figure 1).

Variable	Overall (782)	<12 m (44)	12-23 m (70)	24-59 m (234)	5- < 7 yrs (138)	7- < 10 yrs (296)	p
P. falciparum% (n)	29.5 (231)	22.7 (10)	28.6 (20)	29.5 (69)	27.6 (38)	31.8 (94)	0.5
P. malariae% (n)	0.7 (5)	0.0	2.9 (2)	0.4 (1)	0.0	0.7 (2)	0.4
Gametocytes Pf% (n)	1.5 (12)	0.0	1.4 (1)	0.9 (2)	1.5 (2)	2.4 (7)	0.4
Mean Parasite density par/μl (95% CI)	1,250	765	1770	1410	590	1,380	0.7
	(944-1740)	(179-2920)	(726-4490)	(734-2640)	(299-1590)	(901-2090)
Median haemoglobin, g/dl (IQR)	10.7	11.0	10.6	10.7	10.7	10.6	0.4
	(10.6-10.8)	(10.6-11.4)	(10.2-11.1)	(10.5-11.0)	(10.4-11.0)	(10.4-10.9)
Anaemia (Hb < 11 g/dl)% (n)	51.2 (398)	47.7 (21)	55.7 (39)	47.9 (112)	49.3 (67)	54.1 (159)	0.4

m = months, yrs = years.

p-value used a t-test to compare proportions.

The geometric mean parasite density for P. falciparum was 1,250 parasites/μl (95% CI, 944–1,740) and varied with age, with the highest value in the 1–2 years old group, 1,770 parasites/μl (95% CI, 726–4,490), and the lowest, 590 parasites/μl (95% CI, 299–1,590) in the 5–7 years old group (p = 0.7) (Figure 1). Gametocytes were found in 12 (1.5%, 95% CI, 0.8-2.7) children, all of them P. falciparum, and were more frequent in the older age group (Table 2). No gametocytes were found in infected infants (<1 year old).

The risk of malaria infection was neither different in bed net users (AOR = 0.59, 95% CI, 0.31-3.95, p = 0.22) nor in households sprayed the previous year (OR = 1.23, 95% CI, 0.77-2.01, p = 0.36). The age specific odds show that children 3- < 5 years and those from 5- > 7 years who did not use bed nets the previous night and are in homes not sprayed are more likely to have a malaria infection than those who use bed nets and have their homes sprayed (OR = 6.00, 95% CI, 1.22-29.51, p = 0.03; OR = 4.23, 95% CI, 0.90-19.98, p = 0.06) (Table 3).

This malaria indicator survey, using the World Health Organization (WHO) survey tools, provides a recent estimation of the malaria burden in Nchelenge district. The prevalence of both malaria infection and anaemia was similar to those found in national surveys for Luapula province [13], a result that contrasts with the overall declining trend of malaria in Zambia [14]. Prior national surveys have reported a higher resurgence of malaria in Luapula province than other regions [5]. This could be attributed to seasonality of malaria.

It is important to notice that Nchelenge borders the Democratic Republic of Congo (DRC), with substantial population movements across the borders, and is difficult to reach due to poor terrain, possibly explaining the low coverage of control interventions. Indeed, only a third of the children included in the survey stated to have a bed net and even less used it the previous night. As IRS, only a small proportion of the sampled houses had received this intervention so that any impact on the local vector population is unlikely. It is unknown what the coverage of control interventions may be on the other side of the border but probably not higher than that in Nchelenge, more likely lower. In these conditions, the prevalence of malaria and anaemia found by this survey are not surprising as they reflect the extremely low coverage of ITN and IRS. However, even in case of a better coverage, prevalence may still have been high as the important population movements between Zambia and DRC would maintain the local human reservoir and hence malaria transmission. A significant reduction in the malaria burden in this region may be achieved only through a close collaboration between the national control programmes of both countries. Other countries sharing borders like South Africa, Mozambique and Swaziland in a programme called the Lubombo Spatial Development Initiative (LSDI) have achieved spectacular successes in reducing cross-border malaria burden [15].

The mean parasite density in the children included in this survey, most of them asymptomatic, was relatively high, particularly in the 1–2 years old age group, confirming Nchelenge as an area of intense malaria transmission. Prevalence of malaria would probably have been higher if molecular methods to detect infection had been used. Following continuous exposure to malaria infection, children develop partial immunity and are able to tolerate relatively high parasite densities. Access to treatment seems high as more than 60% of children’s parents stated they had taken an antimalarial treatment the month prior to the survey. However, information on the quality and the source of treatment was not collected during the survey so that it is impossible to know if this was of adequate quality and dosage. The apparently frequent administration of antimalarial treatment is just an indicator of the importance of malaria in this setting.

The prevalence of gametocytes was relatively low though probably a higher number of children with gametocytaemia could have been identified by using molecular methods. Detecting gametocytes in thick blood smears, particularly when present at low densities, is a challenge even for experienced microscopists [16]. Indeed, detecting of P falciparum gametocytes by microscopy can miss most of the gametocyte carriers [17–19] as they are present at an extremely low density though they can still be infectious to mosquitoes [20–22].

This was the weakness of this study in which gametocytes were examined on thick smears.

Coverage by bed nets and IRS was low and this may explain the higher prevalence of malaria infection. However, recent data seem to indicate a problem of emerging resistance to pyrethroids and DDT among the vector population in Nchelenge [23]. Neighbouring countries, such as Mozambique, have found similar patterns of resistance to pyrethroids [24]. If this is confirmed, increasing coverage of both insecticide-treated bed nets and IRS would have a lower than expected impact as ITN would just act as mechanical barriers.

The logistic multivariable analyses found that children did not use bed nets and are in households not sprayed have a higher risk of malaria infection in the age 2 to 7 years. This is the age group that is just weaned from the mother and would not be under a bed net nor cover themselves from mosquito bites. This study also confirms that malaria affects haemoglobin levels in communities where children bear the brunt of the malaria burden [25, 26] and high levels of anaemia are seen as in other parts of Africa as Kenya and Tanzania [27, 28]. Further, children with malaria infection have a higher risk of anaemia, having higher anaemia levels than those without malaria affecting children in many ways including poor performance [29].

The limitations of this work is that it was carried out in relatively smaller or narrow geographical focus. This could over- or underestimate the bed net or IRS coverage and indeed malaria prevalence. Since malaria presents with some seasonality, the study was done at the end of the rainy season but there could be differences in the dry season. Sociodemographic data were collected via self-reporting and, therefore, self-report, recall bias, and social desirability bias cannot be ruled out. Cognisant of this potential bias, the study interviewers made efforts to enhance truthful reporting. The nature of the cross-sectional survey does not permit causal inference of the observed association. In conclusion, even though overall figures indicate that malaria is declining in Zambia, there are still pockets of high endemicity such as the Nchelenge district. This district is on the shores of Lake Mweru and neighbouring DR Congo. Malaria will be controlled only by concerted efforts across borders aiming at increasing coverage of preventive intervention and prompt access to diagnosis and antimalarial treatment.