Non-malaria fevers in a high malaria endemic area of Ghana

The burden of malaria in sub-Saharan Africa is still high [1] although there are areas where it has fallen substantially in the past few years as a result of scaling up of coverage with insecticide treated nets (ITN), selective use of indoor residual spraying (IRS) and introduction of efficacious artemisinin-based combination therapy (ACT) [2‐6]. The relatively recent introduction of rapid diagnostic tests for malaria has shown that a high proportion of fevers in children are due to other diseases, even in settings with high malaria transmission [7]. As a consequence of the reduction in malaria burden, and the wider use of diagnostics, the importance of fevers not due to malaria (non–malaria fevers, (NMFs)) in malaria endemic areas is gaining increasing recognition [8]. The causal agents of non-malaria fevers are diverse, vary from area to area and include protozoa, bacteria and viruses [9‐11].

Some environmental and social factors are known to be associated with NMF. In a study undertaken in Nigeria, the risk of NMF was increased among the poor and among rural populations [12]. Poverty, high population density, poor water supply and climatic changes are associated with specific infections [13‐16]. Another factor identified as a risk factor for NMFs in Benin, West Africa was exposure to malaria antigens in utero, although the mechanism through which this could have been brought about is currently unclear [17, 18].

To investigate the influence of social and environmental factors, placental malaria and gravidity on the risk of NMFs among young children, we have analysed data from the Kintampo Birth Cohort Study which was conducted between 2008 and 2011 in an area of Ghana with high malaria transmission [19]. This study provides further understanding of the epidemiology of NMF in a high malaria transmission area and provides basis of planning health interventions that target NMF.

We followed a cohort of 1855 children during infancy, and investigated the relationship between social and environmental factors, placental malaria and gravidity on the incidence of non-malaria fevers (NMFs). Our findings indicate that despite the well-documented high burden of malaria in this setting, non-malaria fevers contribute a larger fraction of the febrile disease burden in the first year of life.

Since cases of NMF were identified passively and therefore dependent on the health seeking behaviour of mother or caregivers of the infants, it is likely that the true incidence of NMF is higher still. It is possible that children living in this high malaria transmission area are predisposed to other febrile illness due to the modulation of the immune system [28‐30] that results from repeated malaria infections, as reported previously [31, 32].

The high burden of NMFs in young children is concentrated predominantly among poor families living far from health facilities. Children with low birth weight were found to be at an increased risk of NMFs over the course of the study. Low birth weight infants are known to be at a higher risk of bacterial and viral infections [33], possibly due to their immature immune systems as observed among preterm babies [34]. Male children were at a higher risk of having NMF compared to females. Gender differences in disease epidemiology have been well documented among adults and to a lesser extent among children [35]. The risk of bacterial infections and infant mortality are higher in boys than girls [36‐38]. These gender differences may be due to genetic and hormonal differences between boys and girls that may be applicable to our study cohort [39]. However, there may also be differences in care seeking behaviour for illnesses in boys or girls. The incidence of NMF was higher among children born in the long wet season (April–November) compared to children born in the dry season. Birth in the wet season is associated with infections such as respiratory syncytial virus infections [40]. Although we attempted to control for this in a multivariable model, it is also possible that the higher incidence of NMF among infants born during the rainy season may be related to the higher risk of LBW among those born in the wet season.

The finding that NMFs were more frequent in children born to women who had received several courses of IPTp was unexpected. This result was seen in different sub-categories of NMFs, and remained after adjustment for a number of possible confounders including distance from a health centre, place of residence and SES. It also remained in exploratory analyses over a longer period of follow-up. One possibility is that mothers who received IPTp were especially conscientious about seeking care for their children when they were ill, resulting in higher apparent incidence (i.e. the apparent link between higher IPTp use and NMF incidence is a result of confounding by care-seeking/access to care). However, if this was the case, it might also be expected that the children of these mothers would have an increased incidence of malaria; this was not found in our previous analysis of children in this cohort [19]. It is difficult to suggest a mechanism by which IPTp could increase the risk of non-malaria fevers, without increasing risk of malaria and consequently this may be a chance finding, or attributable to residual confounding; this should be examined in future studies.

NMFs were no more frequent in children born to women with placental malaria at delivery than in children whose mothers did not have placental malaria either in crude or adjusted analyses. There was also no association between the incidence of NMFs and mothers’ gravidity. In a previous study carried out in Benin, placental malaria infection was associated with a higher risk of NMFs among young children [18] and the investigators suggested that exposure to malaria antigens in utero predisposes young infants to NMF [18], as has been suggested previously for malaria [17]. The increased incidence of malaria among infants born to mothers with placental malaria (PM) in relatively low malaria transmission areas [41‐45] could be due to exposure of these infants to malaria antigens in utero leading to changes in the immune system which make them more susceptible to malaria [17]. Alternatively, mothers with placental malaria may be at increased risk of malaria due to a combination of social and environmental risks which may be shared with their infants [19]. These social and environment factors could also increase the risk of NMFs in infants born to women who had placental malaria.

The reason for differences between our results and those of the study in Benin is unknown, although we performed analyses to match those undertaken in the Benin study as closely as possible. There were however some methodological differences between the two studies. Firstly, the transmission of malaria in Kintampo was much higher than in Tori Bossito, Benin (entomological inoculation rate (EIR) 269 infectious bites per person per year [20] and 20.5 respectively) [42, 43]. Thus, levels of in utero exposure to malaria may be higher in Kintampo than in Tori Bossito. Secondly, we used placental histology to diagnose placental malaria, a more sensitive test for placental malaria than the use of placental blood smears [46‐48], the technique employed in Benin. Therefore we may have detected placental infections in this study that would have been missed in the previous study [18]. This could have biased results in the earlier study if high density infections acquired more recently were more likely to be detected, since this would tend to identify women with higher exposure to malaria (who may therefore be poorer, live in rural areas, etc., and therefore be more susceptible to malaria and NMFs). Finally, more children were followed in the present study than in the study undertaken in Benin. Although it was not possible to observe all study children until 18 months of age, exploratory analyses restricted to those who were followed to 18 months of age and beyond indicated very similar results to our analyses focusing on infancy.

The following potential biases are acknowledged in this study. Firstly, NMF cases were passively detected, therefore the true incidence may have been under-estimated depending on the health seeking behaviour of caregivers. To address this, care-givers were provided with free transportation and health insurance, and encouraged to attend study clinics whenever their child was unwell. Secondly, since there is no diagnostic test routinely carried out for the non-malarial causes of fever, it is possible that co-infections of malaria and other causes of fever such as viral infections were misclassified as malaria-only infections. This may also have led to underestimation of the incidence of NMF. Similarly the incidence of NMF may be underestimated if mortality among the study participants were as a result of NMF that were not reported.

Despite these issues that may lead to an underestimate of malaria incidence, and despite the high level of malaria transmission in this area of Ghana, the incidence of NMF is high. This underscores the importance of other causes of fever in many settings where malaria has previously been the major focus. There is an urgent need to determine the most frequent causes of NMFs in different ecological situations to allow the development of therapeutic and preventative strategies based on sound scientific evidence.

The incidence of NMF is high in the middle belt of Ghana. The incidence of NMF is associated with low birth weight and poor socioeconomic status but not with placental malaria.

ACT, artemisinin-based combination therapy; aHR, adjusted hazard ratio; CI, confidence interval; EIR, entomological inoculation rate; IPTp, intermittent preventive treatment in pregnancy; IRS, indoor residual spraying; ITN, Insecticide treated nets; KHDSS, Kintampo Health and Demographic Surveillance System; KHRC, Kintampo Health Research Centre; LBW, low birth weight; MG, multigravidae; NMF, non malaria fevers; PG, primigravidae; PM, placental malaria; SES, socio-economic status; uHR, unadjusted hazard ratio