Malaria causes more than 500 million clinical cases and is responsible for about one million deaths annually, mostly among African infants and young children [1]. Uganda bears a particularly large burden, having among the highest rates of transmission worldwide [2]. Within Uganda, there is significant heterogeneity of malaria transmission. In the most recent malaria indicator survey, from 2009, parasite prevalence among children under five years of age ranged from <5% in urban centres to >60% in rural settings [3].

As utilization of malaria control interventions, including long-lasting insecticide treated bed nets (LLINs), indoor residual spraying of insecticides, and prompt treatment with artemisinin-based combination therapy (ACT) has increased, many reports have shown a substantial drop in malaria transmission, malaria-associated hospitalizations, and malaria-associated deaths [1,4-8]. However, these findings have not been uniform across Africa [5,9,10]. One recent study showed a lack of decline in slide positivity rates and admissions for cerebral malaria from 2001–2010 in Malawi, and another showed an increase in malaria-related hospitalizations over a similar time period in Uganda [9,11]. These, and many other relevant recent studies, were retrospective in design, and so unable to provide reliable estimates of malaria incidence.

Considerable knowledge regarding the natural history of malaria, acquisition of anti-malarial immunity, and impact of interventions has been gained from longitudinal cohort studies. Beginning with Koch’s seminal studies in Papua New Guinea in the late 1800s, these studies have consistently revealed that, in endemic populations, the incidence and severity of malaria decreases considerably after the first years of life, with a corresponding rise in the prevalence of asymptomatic carriage of parasites [12-15]. Although widespread use of LLINs and ACT have been associated with declines in the incidence of malaria in many regions, there have been concerns over how decreasing exposure to malaria parasites may alter the natural history of disease, and possibly delay the acquisition of anti-malarial immunity [5,10,15]. To better understand the impact of LLINs and ACT on the natural history of malaria in a high endemicity setting, the incidence of malaria and prevalence of asymptomatic parasitaemia were evaluated in a cohort of children living in Tororo, Uganda.

In this longitudinal cohort of 100 children given LLINs and ACT and residing in an area of known high transmission intensity in eastern Uganda, a remarkably high incidence of malaria was observed, peaking at 6.5 episodes per child per year at 25 months of age. Importantly, the vast majority (99.7%) of malaria cases were uncomplicated. The incidence declined as children reached three years of age, suggesting the development of natural immunity to malaria [23], although children on average still had four episodes per year at four years of age, and asymptomatic parasitaemia, a finding typically associated with anti-malarial immunity, was observed in less than 10% of individuals. Perhaps most concerning, in the setting of near universal use of LLINs to limit mosquito exposure and ACT to effectively treat malaria, the incidence of malaria was very high throughout this study and rose 52% from 2008 through 2011.

Many reports across Africa have shown significant declines in malaria-related deaths and hospitalizations over the past decade, and it is routinely suggested that malaria-control interventions - including usage of LLINs and ACT - are responsible for these declines [4-8]. However, few of these studies were done with longitudinal cohorts in high-endemicity settings. Contrasting with many reports, this study and recent reports from Malawi, Uganda, and Senegal showed a lack of decline in malaria incidence [5,9,10].

The reason for a lack of decline, and indeed a significant increase in the incidence of malaria in Tororo is unknown. The increase may be due to changes in drug resistance, vector species, insecticide resistance, and/or climate/rainfall, leading to increases in exposure to malaria vectors. Considering drugs, AL and DP have both shown outstanding efficacy in numerous recent studies, and there is no convincing evidence of resistance to these agents in Africa. Although we did not collect detailed information on mosquito species or biting habits, preliminary data from the East African International Centers of Excellence in Malaria Research (ICEMR) estimates the EIR in Tororo to be 379 infective bites PPY (unpublished data from CDC light traps collected between Oct 2011 and Sep 2012), near the EIR of 562 infective bites PPY estimated by human landing catches in 2002 [16]. The majority of mosquitoes collected during this period were Anopheles gambiae complex (93.5%), as found in 2002 [16]. Notably, population based surveys in this area have revealed significant increases in LLIN coverage in Tororo over the past six years, with the proportion of children <5 years of age sleeping under a LLIN increasing from 13% in 2006 [24] to 42% in 2009 [25] and 62% in 2012 (East African ICEMR, unpublished). However, several recent reports have also shown increasing resistance of anopheline mosquitoes to pyrethroid insecticides in Tororo. Prevalence of the knockdown resistance (KDR) L1014S mutation has increased from 29% in 2002 to 75% in 2008 in mosquitoes collected from Tororo [26,27], suggesting that the effectiveness of LLINs may be waning in this area, as also seen in other parts of Africa [10,28]. In these settings, additional control measures - including indoor residual spraying [29], larvicides and other vector control measures [30], and chemoprevention [31], will be required.

Importantly, even though more than 1,600 cases of malaria were diagnosed in the 100 children enrolled in this cohort, remarkably few cases of malaria were associated with complications. Only six episodes met criteria for complicated malaria, and all these cases were due to a single convulsion (danger sign) and did not meet criteria for severe malaria. There were no cases of cerebral malaria or respiratory distress. This result is in sharp contrast to historical reports suggesting that 2% of all clinical attacks of malaria are severe [32]. Consistent with earlier observations from a cohort study in Uganda conducted in a lower transmission setting [33,34], these findings suggest that malaria-related morbidity can be greatly limited with prompt access to appropriate diagnosis and highly effective treatment. Early treatment failures were also exceedingly rare, consistent with prior findings that ACT is very efficacious [35-38]. Of 12 treatment failures within 14 days of prior therapy, there were three cases of severe anaemia, all in the same child. The unfortunate death of this two-year old boy, who had had 27 prior episodes of malaria, is clear evidence that, even with prompt diagnosis and therapy, malaria remains a dangerous disease in need of improved control measures.

As others have reported, children in this cohort appeared to acquire anti-disease immunity as they aged. The incidence of malaria declined in children after 30–36 months of age, and asymptomatic parasitaemia was more common in older children [23]. However, given the degree of exposure and high incidence in the cohort, the natural acquisition of anti-malarial immunity in this cohort was remarkably slow. A surprising finding was the low prevalence of asymptomatic parasitaemia in this cohort - 5% based on monthly microscopy. In contrast, parasite prevalence rates from eastern Uganda in the 2009 malaria indicator survey were nearly 40% in children <5 [25], and in a similarly aged cohort from Kampala, Uganda, a lower transmission setting, the prevalence of asymptomatic parasitaemia was much higher (17%) [39]. Given a lack of standard diagnostic criteria, asymptomatic parasitaemia described in this report accounted for whether or not children were in a pre-symptomatic period, excluding any child that developed malaria within seven days [40]. Although varying definitions may explain some of these differences, the significantly low prevalence of asymptomatic infection observed here suggests that prompt and repeated treatments with highly effective ACT may alter the acquisition of anti-parasite immunity, limiting infections to symptomatic disease in young children. Further studies addressing the impact of ACT on the development of anti-malarial immunity are needed.

There were several limitations to this study. The cohort was enrolled by convenience sampling, limiting the generalizability of the findings. Although the analysis attempted to adjust for calendar time, children enrolled early in the study may have differed from those enrolled later. Detailed information was also unavailable regarding individual- and household-level exposure. A further limitation was the assessment for LLIN use by self-report, and not by directly observed LLIN usage, which may have overestimated LLIN coverage [41]. Finally, children received treatment for malaria if they presented with fever and any parasitaemia. If non-malarial causes of fever were the aetiology of some of these presentations, malaria incidence may have been overestimated in this cohort. However, the geometric mean parasite densities of malaria episodes were significantly higher than those of episodes of asymptomatic parasitaemia, suggesting that malaria contributed to symptoms in most cases where anti-malarial therapy was given.

In conclusion, despite reports of decreasing malaria morbidity and mortality across many parts of Africa, the incidence of malaria continues to be very high in Tororo, Uganda, even in a setting with LLINs and ACT, and this incidence appears to be rising. Thus, additional malaria control interventions among young children living in high transmission settings are needed. Fortunately, with prompt diagnosis and effective treatment, malaria-related complications can be limited even in high transmission settings.

The authors declare that they have no competing interests.

GD, JT, PJR, and MRK conceived and designed the study. MKM, AK, and EA participated in data collection. PJ, BG, and GD participated in the data analysis. All authors participated in the writing of the manuscript, read and approved the final manuscript.