With helminths and malaria infections endemic through most of Africa, communities often endure infections with a number of different parasite species [5], and individuals are often ‘co-infected’ with combinations of helminths, and malaria parasites [6, 7]. Rates of co-infection may not only depend on chance, but also upon the spatial distribution of environmental conditions that favour transmission of multiple species [8], as well as upon immunological interactions and common factors in genetic susceptibility or host behaviour.

Immunological factors are expected to influence rates of co-infection because helminths modulate host immune responses both to themselves, and to concurrent infections [9-11]. With regard to malaria, murine models provide evidence of interactions manifesting in altered probability of morbidity or mortality[12-16]. For example, infection with the filarial nematode Brugia pahangi has been found to protect against development of cerebral malaria [14], while super-infection with Schistosoma mansoni delayed clearance and increased severity of Plasmodium infection [15]. Among humans, published studies have not led to any consensus [17-24]. Reports from Thailand suggest that hookworm, Trichuris trichiura and Ascaris lumbricoides infections may increase the incidence of Plasmodium falciparum infections [18] whilst also suggesting that A. lumbricoides infection protects against severe malaria [19, 20]. This agrees with an earlier report and a recent trial suggesting that the incidence of malaria attacks increased after treatment of severe A. lumbricoides infection[21, 25], but studies elsewhere have not supported these findings [16, 17, 22]. Inconsistent results have also been reported in studies of schistosome and malaria co-infections [23, 24]. These contrasting results leave important unanswered questions about the biological associations between malaria and helminths.

These contrasting results may be explained by a lack of assessment of spatial variation in exposure to parasitic infections [8]. Parasites are subject to micro-geographical variation in the risk of infection: for example, schistosomiasis is found around water contact sites, risk of malaria parasite infection is affected by distance to a larval breeding source [26], and A. lumbricoides, T. Trichiura and hookworm are influenced by environmental conditions [27]. As exposure to multiple species of parasitic infection may vary over small distances, there is a clear need for analyses to consider residential location as a confounder of the risk of co-infection: something that few previous studies have attempted [8].

Pregnant women are an under-studied group, but are more vulnerable to infections, due to suppression of the immune system during pregnancy [28, 29]. If helminths do have a biological effect on susceptibility to malaria, this may be particularly important in pregnancy, where malaria is associated with increased maternal mortality and anaemia, IUGR and foetal and perinatal death [30], and maternal immune responses influence offspring’s immunity and response to malaria infection in infancy [31].

The study design and selection criteria have been described previously [32]. Women were enrolled at Entebbe General Hospital antenatal clinic between April 2003 and November 2005. They were eligible if they were pregnant at the time of enrolment, resident in Entebbe or Katabi, and in good health. They were excluded if they were severely anaemic (haemoglobin <8 g/dl), the pregnancy was not normal, they were unwilling to receive an HIV test result (as part of the hospital programme for Prevention of Mother To Child Transmission of HIV), or if they were unwell on the day of enrolment. Recruitment took place over two visits to the clinic; ‘screening’, and ‘enrolment’ visits.

At the screening visit, women gave blood for examination for malaria parasites and M.perstans. Socio-economic and demographic data were collected by questionnaire. Women returned with a stool sample for the enrolment visit. All samples were collected as part of the baseline survey before treatment was given.

Stool samples were examined using the Kato-Katz method [33], and charcoal culture for Strongyloides stercoralis [34]. Two Kato-Katz slides were prepared for each sample and examined within 30 minutes for hookworm, and the following day for other parasites. The intensity of hookworm infection was categorised as follows: light <1,000 eggs/gram of stool (epg), moderate 1,000-3,999 epg, high ≥4,000 epg. [35]

Blood was examined using a thick film for malaria parasites and modified Knott’s method for Mansonella perstans with intensity estimated as malaria parasites per 200 white blood cells, and microfilariae per ml blood [36].

The residence of each participant was georeferenced using Garmin handheld Global Positioning System units, during a survey carried out in the first quarter of 2006. If the participant was found to have moved since enrolment, her address at enrolment was visited and georeferenced.

Approval for the study was given by of the Uganda Virus Research Institute Science and Ethics Committee; the Uganda National Council for Science and Technology; and the London School of Hygiene and Tropical Medicine.

Analyses were conducted using STATA version 7 (College Station, Texas, USA). ArcGIS Desktop (Environmental Systems Research Institute, California, USA) was used to assign subjects to geographically defined zones on the basis of their coordinates. The analysis was divided into two parts. First was the analysis of co-infection by helminth species and P.falciparum, adjusting for maternal demographic, socioeconomic and clinical confounding factors, and second was a more detailed assessment of co-infection in relation to geographical area.

Simple univariate and adjusted analysis of association with P.falciparum was performed for each helminth species using logistic regression. Variables considered as potential confounding factors, and included in the initial model, were age, tribe, woman’s socioeconomic status, household socioeconomic status, geographical zone and HIV status. Two scales for socio-economic status, each with six levels of scoring, were devised from the questionnaire. ‘Woman’s Socioeconomic Status’ was determined by the woman’s level of education, personal income and occupation. ‘Household Socioeconomic Status’ was determined by building materials, number of rooms and items owned [37].

Initial crude analyses showed a strong positive association between hookworm and malaria and between M.perstans and P.falciparum. There were no statistically significant associations between S.mansoni, T.Trichiura or S.stercoralis and P.falciparum (Table 1). The strength of the association increased with intensity of infection for hookworm (odds ratio (OR) for malaria compared to individuals without hookworm: 1.43 for light hookworm infections, 2.14 for moderate and 2.36 for heavy infections (test for trend, p<0.001)). No such trend was observed for intensity of M.perstans infection. Hookworm, M.perstans and P.falciparum infection were significantly associated with age and socioeconomic status; hookworm and M.perstans were associated with tribe; hookworm and P.falciparum were associated with HIV infection (data not shown). After adjusting for these potential confounding factors, the association between hookworm and P.falciparum was reduced. The association between M.perstans and P.falciparum was reduced, but remained strong.

There was also a statistically significant association between hookworm and M.perstans (OR 2.70 (95% confidence interval (CI) 2.20-3.31, p=<0.001)) and an interaction between hookworm and M.perstans in relation to their associations with P.falciparum (p=0.047): the prevalence of P.falciparum was 7.5% percent among participants with neither hookworm nor M.perstans, 11.5% among those with hookworm only, 18.8% among those with M.perstans only and 17.2% among those with both. Thus the hookworm-P.falciparum association was seen only in the absence of M.perstans. Adding the interaction term to the model increased the strength of the individual associations with P.falciparum for hookworm and M.perstans (aOR: for hookworm-P.falciparum in the absence of M.perstans 1.43 (95% CI 1.03-1.98; p=0.034); for M.perstans-P.falciparum in the absence of hookworm, 2.29 (95% CI 1.46-3.59; p<0.001) and for both hookworm and M.perstans with P.falciparum, 1.80 (95%CI 1.23-2.65; p=0.003).

The geographical distribution of women’s homes and of zones is shown in the Figure and the distribution of infections by zone is indicated in table 2. There was an absence of P.falciparum infection in the area to the extreme southwest of the peninsula (zone one), and in the urban area of Entebbe town centre to the east of the airport runway (zone two). Two areas with increased density of infection were found to the northwest of the town; one at the northwest tip of the spur that reaches the inlet of the lake (zones three and four), and one on the edge of the adjacent swamp (zone three). Moving inland to the north was a band of more diffuse P.falciparum infection. Hookworm infection was spread throughout the study area, with little geographical clustering. M.perstans infection was spread across the study area with increased infection in the easternmost spur (zone 12). Moving north and west from this area were two further areas with increased density of infection.

In addition to prevalence of P.falciparum, hookworm and M.perstans infections, the adjusted ORs for co-infection, stratified by zone (defined in the Figure), are shown in Table 2. The adjusted ORs for hookworm-P.falciparum co-infection varied by geographical location: for example, in zone nine there was a particularly strong association, not observed in zones four, seven and eight. The association between M.perstans and P.falciparum infection was more consistent.

This study focused on the issue that the geographical distribution of parasitic infections exhibit spatial dependency over small distances [27], and represents a step forward from studies conducted in one location that had assumed no spatial clustering, or had not measured residential location[18, 22, 38]. Geographical zones were defined based on simple geography, altitude, vegetation and location of settlements, as these have been found to correlate spatially with parasite infection [27, 39]. The zones provide a means to analyse the different environments of, for example an exposed costal location of zone 1 compared with the very different urban environment of zone 3 as separate entities, and to adjust for these differences in analysis. However, the stratification had limitations; while the aim was to provide a detailed stratification by environment, there was a need to strike a balance between achieving homogeneity within the zones without creating zones so small that the analysis is not sufficiently powered. With probability of infections known to exhibit spatial dependency over small distances [27], it may have been incorrect to assume homogenous parasite density within the study zones of up to four km in diameter. The use of zones in the analysis presented here enabled the description of variation between areas, but not within, areas..

Stratification by zone revealed considerable variation in the probability of co-infection with geographical location, particularly for hookworm-P.falciparum. This may be related, in part, to the observed variability in the prevalence of hookworm and P.falciparum infections. Associations were strongest in zones where the infection prevalence of both hookworm and P.falciparum were highest, consistent with a hypothesis that probability of infection with P.falciparum increases with intensity of hookworm transmission.

A considerable number of women seen at the antenatal clinic were excluded from this study; it is unlikely that the major reasons for exclusion created an important bias. The principal reason for exclusion, residence outside the study area, was appropriate to this analysis. A bias in relation to HIV status of those excluded is unlikely to have affected the results, since the principal effects showed no interaction with HIV among those analysed.

Women were included in this study only if they were well on the screening day, with no complaints (e.g. fever) and no gross evidence of severe, helminth-induced disease (such as anaemia, bloody diarrhoea, or overt liver disease). This analysis therefore addresses associations between helminth infections and asymptomatic P.falciparum parasitaemia, which may differ from associations between helminths and symptomatic P.falciparum infection [40]. Assuming that helminth infections are more often chronic and long-lived than infection with P.falciparum, the observed positive association between helminths and asymptomatic P.falciparum parasitaemia may imply an increased likelihood of P.falciparum infection, a reduced likelihood of clearing P.falciparum, and, or, a reduced likelihood of developing symptoms and seeking medication.

The first possibility, a helminth-P.falciparum association due to increased likelihood of infection with P.falciparum among women with helminth infections, could arise through behavioural or environmental factors, leading to increased exposure to both types of infection. Both M.perstans and P.falciparum are transmitted by flying insect vectors: M.perstans by Culicoides midges [41] and P.falciparum by Anopheles mosquitoes [26]. It is plausible that the distribution of these two vectors may be spatially correlated, as both require water sources for larval breeding [26, 42] though it is not clear whether the required conditions are exactly the same. Similarly, hookworm larvae flourish in damp soil and grass, which may be found close to stagnant water that is the breeding ground for malaria parasites. The slight reduction in the helminth-P.falciparum associations with adjustment for socio-demographic factors suggests a possible contribution of other behavioural effects; such as differences in the usage of antimalarial drugs during pregnancy, prior to enrolment in the study; however, prior consumption of antimalarials showed no statistically significant association with helminth infection and adjustment for prior consumption of antimalarials did not alter the observed effects (data not shown). On the other hand, adjustment for geographical zone strengthened the associations, suggesting that, at zone-level, common environmental factors could not explain the effect. Thus the possibility that helminths may lead to a biological increase in susceptibility to infection, or to persistence of asymptomatic infection with P.falciparum remains plausible. This accords with previous studies suggesting associations between helminths and an increased incidence of P.falciparum infection [20], higher parasite counts and delayed parasite clearance [17], but a reduction in disease severity [21-24]. It is not possible to say whether the higher prevalence of P.falciparum among helminth-infected women resulted in increased incidence of disease events from the data in this cross sectional study.

Hookworm and M.perstans were the commonest helminth infections among participants in this study. The less common infections, S.mansoni, Strongyloides, Trichuris and Ascaris, showed no associations with P.falciparum, in conflict with some previous reports [20, 43-45], but in agreement with one other study from Uganda that reported no association [17]. For these species, a real association (if present) might not have been detected in this study because few women were infected (particularly with Ascaris), because the intensity of helminth infections was low, or because of misclassification of low-intensity infections as negative, resulting from the examination of a single stool sample (multiple samples are required for high sensitivity of detection) [46-48]. Such misclassification may also have contributed to the relatively weak observed effect of hookworm. By contrast, preliminary results suggest that the Knott’s method used for assessment of M.perstans infection was particularly robust, with serial results in the same women showing 96% agreement for infection status and a correlation coefficient for microfilarial counts per millilitre of 0.88 (p<0.001) (AME, unpublished data).

This study does not explore potential biological mechanisms for the observed associations between helminths and P.falciparum. However, previously proposed mechanisms include the suggestion that the immuno-regulatory effects of helminths, which allow their own long-term survival in the host [9], “spill-over” to impair the immune response required to protect against or eliminate malaria parasites. M.perstans is a long-lived filarial worm that inhabits serosal body cavities and reproduces through microfilaria which circulate in the blood and are transmitted through biting midges [42]. Despite residence and migration through blood and tissues, M.perstans infection seldom causes detectable pathology, and this attests to its particularly potent immuno-modulating properties. It is thus interesting to speculate that M.perstans might produce a particularly strong immuno-modulating effect on the response to other pathogens, and that this might over-ride the effects of related helminths, such as hookworm, when both are present, perhaps accounting for the observed interaction between the two helminth species. This will be more comprehensively described in a future paper.

This study specifically examined co-infections in pregnant women. The unique environment that exists within the pregnant body means that one should be cautious in applying these results to the general population. This may be particularly important in relation to associations with P.falciparum, because parasites are sequestered within the placenta during pregnancy [30] and may be less readily detected in the peripheral blood (sampled in this study). Apparent biological associations might possibly reflect helminth effects on the sequestration of P.falciparum parasites in the placenta, rather than effects on the prevalence of infection.