Seasonality in malaria transmission: implications for case-management with long-acting artemisinin combination therapy in sub-Saharan Africa
- 2.7k Downloads
Long-acting artemisinin-based combination therapy (LACT) offers the potential to prevent recurrent malaria attacks in highly exposed children. However, it is not clear where this advantage will be most important, and deployment of these drugs is not rationalized on this basis.
To understand where post-treatment prophylaxis would be most beneficial, the relationship between seasonality, transmission intensity and the interval between malaria episodes was explored using data from six cohort studies in West Africa and an individual-based malaria transmission model. The total number of recurrent malaria cases per 1000 child-years at risk, and the fraction of the total annual burden that this represents were estimated for sub-Saharan Africa.
In settings where prevalence is less than 10 %, repeat malaria episodes constitute a small fraction of the total burden, and few repeat episodes occur within the window of protection provided by currently available drugs. However, in higher transmission settings, and particularly in high transmission settings with highly seasonal transmission, repeat malaria becomes increasingly important, with up to 20 % of the total clinical burden in children estimated to be due to repeat episodes within 4 weeks of a prior attack.
At a given level of transmission intensity and annual incidence, the concentration of repeat malaria episodes in time, and consequently the protection from LACT is highest in the most seasonal areas. As a result, the degree of seasonality, in addition to the overall intensity of transmission, should be considered by policy makers when deciding between ACT that differ in their duration of post-treatment prophylaxis.
KeywordsMalaria epidemiology Seasonality Heterogeneity Artemisinin-based combination therapy Post-treatment prophylaxis
artemisinin-based combination therapy
long-acting artemisinin-based combination therapy
larval carrying capacity
Markham seasonality index
polymerase chain reaction
seasonal malaria chemoprevention
The burden of malaria is not shared equally in endemic areas—some children experience repeated attacks of malaria while others remain healthy [1, 2]. Children who have multiple episodes of clinical malaria may be more likely to develop severe disease or severe anaemia [2, 3, 4, 5], and recurrent malaria after discharge may lead to further severe morbidity in children who have been admitted to hospital [6, 7]. Repeated episodes of malaria may make children more susceptible to, or increase the severity of, other conditions such as invasive bacteraemia , and may also lead to poor school performance .
Artemisinin-based combination therapy (ACT) used for case management of malaria provides protection against a subsequent attack in the immediate period after treatment, because parasites that emerge from the liver while sufficient concentrations of the partner drug remain present in the blood will be killed. The duration of this post-treatment prophylaxis (PTP) depends on the pharmacokinetics and pharmacodynamics of the drug, as well as the resistance patterns in the parasite population to which individuals are exposed . Thus, some repeat episodes of malaria will be prevented by use of any effective drug, but the number of episodes prevented will be greater where a longer-acting drug is used to manage the initial episode  and where the interval between attacks is short. Several long-acting forms of ACT (LACT) are now available, including dihydroartemisinin-piperaquine (DHA-PQ) and artesunate-mefloquine (AS-MQ), which provide several additional weeks of PTP compared to shorter-acting drugs. However, choice of first line treatment is not currently rationalized to exploit this benefit to its full potential.
However, to explore this further, and to generalize these findings to other drug combinations and to other interventions that could potentially be deployed to reduce the burden of repeated episodes of malaria, a better understanding of how seasonality and transmission intensity affect the epidemiology of repeated episodes of malaria is needed. This information could help inform policy-makers as to the likely importance of repeated episodes of malaria within a specific context, based on these underlying epidemiological drivers. In the present study, the timing and interval between repeat malaria episodes were estimated using data from six cohort studies in West Africa. An individual-based model of malaria transmission across sub-Saharan Africa was then used to explore the effects of transmission intensity and seasonality on the timing of repeat malaria across a range of epidemiological settings.
Details of the cohort studies in West Africa
No. malaria episodesa
Year of study
Drug used for case-management
Child age group
Insecticide-treated net coverage (%)
Chloroquine (first line), sulfadoxine-pyrimethamine (2nd line)
Under 5 years
2002: 36 % pre-season, 37 % post-season; 2003: 31 % pre-season, 30 % post-season
Farafenni, The Gambia
Chloroquine plus sulfadoxine-pyrimethamine
Under 10 years
post-season 2003: 38.7 % (under 10), 36.3 % (under 5)
10–40 (1997) 
Kati district, Mali
Under 5 years
13.2 % at end of transmission season
6.6–37.3  (supplement S3)
Boussé district, Burkina Faso
Under 5 years
41.5 % at end of transmission season
31.5 % at 18 months of age
Perennial, peaks April–October
64 % (under 5) 
231–269 (2003–2005) 
The Markham seasonality index (MSI)  has previously been used to characterize malaria seasonality . Full details are given in Additional file 2. In brief, the malaria incidence in each month is expressed as a vector, with a (different) fixed direction for each month, and with the length of the vector corresponding to the fraction of the annual incidence occurring in that month. Summation of the 12 monthly vectors gives a resultant vector, i.e. the finishing point back to the origin. The length of the resultant vector divided by the total length of all 12 monthly vectors gives the MSI (taking values ranging from 0, in the case where all months have equal incidence, and 1, in the case where all incidence occurs within 1 month). The direction of the resultant vector indicates the month in which the peak in incidence occurs. The MSI was calculated for each of the six studies.
Timing of repeated episodes
The distribution of recurrent malaria episodes was inspected visually by constructing a histogram of the interval between episodes experienced by the same individual (based on date of first contact for each unique malaria episode). The mean interval between episodes was then obtained using the Kaplan–Meier estimate of the mean survival time among those who experienced a subsequent episode before the end of the follow-up period.
Having calculated these measures of seasonality and distribution of intervals between of repeat episodes for the six sites, an individual-based model (IBM) of Plasmodium falciparum described previously  was used to generalize these findings. This model describes the full transmission cycle of P. falciparum between humans and mosquitoes, as well as disease progression in humans, and has been fitted to extensive data on parasite prevalence determined by microscopy and polymerase chain reaction (PCR), and episodes of uncomplicated malaria by age and transmission setting across Africa [25, 26]. Heterogeneity in malaria risk is simulated by randomly assigning individuals at birth to experience different relative biting rates . The model also incorporates the impact of insecticide-treated nets (ITNs). Anti-disease immunity, which reduces the probability of a blood-stage infection resulting in a clinical episode, is acquired at a rate that depends on exposure [25, 27]. The assumed parametric form of this immunity in the model gives rise to a lengthening of the interval between successive malaria attacks in areas of high endemicity, particularly among the most exposed sub-groups within a population who have higher exposure, and who therefore acquire protective immunity faster relative to their neighbours. The model has been fitted to age-incidence curves from 23 sites in Africa  but has not previously been compared to longitudinal patterns of repeat episodes. For this analysis, the timing of clinical episodes of malaria predicted by the model (i.e. when infection is symptomatic, and treatment is sought) were recorded for each individual in the simulation, to create a dataset of the timing of treated clinical episodes comparable to those from the six studies in West Africa.
Simulations were run using transmission and seasonality parameters appropriate to the first administrative unit (the largest sub-division within each country) within which the studies were located, to investigate if the model could approximately replicate the patterns seen in the six data sets in terms of seasonality and concentration of malaria episodes. Duration of post-treatment prophylaxis after treatment for malaria in each simulation was chosen to approximate that provided by the drug used in each study. Simulations were restricted to the same age group and with prevalence and levels of ITN coverage similar to those in the original studies.
Seasonality in transmission in the model is driven by temporal fluctuations in rainfall and temperature, which determine the ability of the environment to support development of mosquito larvae, referred to as larval carrying capacity (LCC) . The LCC function was integrated by month for each of the 576 first administrative areas across Africa to estimate the MSI as described above, giving values of the MSI ranging from 1 to 91 %. However, all sites with an MSI <10 % were bimodal sites with two equal peaks spaced approximately 6 months apart. Nine sites were chosen representing each 10 % interval of the MSI between 10 and 90 % for further analysis, focusing on those without bimodal seasonality patterns, which were considered separately (see Additional file 3). To confirm that the seasonality in larval carrying capacity was reflected in incidence of clinical malaria predicted by the model, the Markham seasonality index was also calculated for symptomatic attacks of malaria predicted by the model in the nine sites.
Modelling seasonality in different transmission intensity settings
For each of the seasonality patterns identified as described above (i.e. with MSI between 10 and 90 %), scenarios were simulated with mean annual parasite prevalence in 2–10 year old children of 5, 10, 20, 40 and 60 %, creating a 9 × 5 matrix of seasonality versus transmission intensity. For each of the 45 scenarios, all other parameters apart from the seasonality pattern and prevalence in 2–10 year olds were held constant. The set of parameters from the Upper East Region of Ghana (the first administrative unit in which the Navrongo study site is located) were used. This site was chosen because the epidemiology in this location has been well-characterized e.g. [29, 30, 31, 32, 33, 34, 35] and because this site was one of a number of sentinel sites used in assessing the fit of the model . For all scenarios in the simulation, duration of post-treatment prophylaxis after case-management for malaria was assumed to be short (mean duration of 10 days).
Histograms showing the distribution of intervals between episodes were produced, and the estimated mean interval between malaria episodes calculated for each of the modelled scenarios using the Kaplan–Meier approach as described above. To understand the relative importance of seasonality and endemicity on the timing of repeat malaria episodes, the mean interval between episodes was plotted against the Markham seasonality index for each level of prevalence.
To explore the possible effect of post-treatment prophylaxis of varying duration, the model was used to estimate, for each scenario, the number of repeat malaria episodes that occur within 28, 42, 56 and 70 days of a previous attack, and the fraction of the annual burden that this represents. The result for the corresponding model scenario was then matched (in terms of seasonality and prevalence category) to indicate the absolute and relative importance of repeat malaria in these different periods, for each first administrative area in Africa.
Data from cohort studies
In the cohort studies in West Africa, the seasonality in clinical malaria episodes increased moving from South to North, following the known patterns in seasonality of rainfall in the Sahel and sub-Sahel . The Markham seasonality index (MSI), used to quantify seasonality in malaria incidence, was as follows: Kintampo, Ghana : 42.8 %; Navrongo, Ghana : 54.5 %; Boussé, Burkina Faso [18, 19]: 74.3 %; Kati, Mali [20, 21]: 82.3 %, Farafenni, The Gambia (Milligan, unpublished data): 85.5 %, Niakhar, Senegal : 79.0 %. A graphical representation of the Markham seasonality index is shown for each site in Additional file 4.
Interventions able to tackle the burden of repeat malaria soon after treatment are likely to have an important impact on malaria burden in highly endemic settings, and particularly in highly endemic and highly seasonal settings. In such settings, repeat malaria within 28 days of a previous episode constitutes up to 20 % of the overall malaria burden in children (Additional file 9). Within 42 days, this share of the overall burden rises to around 30 %. At 60 % prevalence, an estimated 250 cases per 1000 children per year occur within 42 days of a prior attack in perennial settings, rising to 1000 cases per 1000 children in highly seasonal settings (Additional file 10). In such areas, the incremental effect of a drug providing an additional 2 weeks of PTP (42 days versus 28 days) could prevent in excess of 200 cases per 1000 children, per year.
In areas with highly seasonal transmission and moderate-high transmission (e.g. 2–10 year old prevalence of 40 %), repeat malaria can, relative to non-seasonal settings with substantially higher prevalence (2–10 year old prevalence of 60 %), account for a higher number of cases (e.g. Additional file 10) and a higher proportion of the overall malaria burden (Additional file 9).
In settings with a prevalence in 2–10 year old children less than 20 %, repeat malaria is relatively unimportant. This is partly because there is a small absolute number of repeat episodes (Additional file 10), and partly because repeat episodes soon after a previous attack constitute a small fraction of the overall burden (Additional file 9).
While the burden within 56 and 70 days of a prior episode is larger than that within 28 and 42 days, the incremental increase in the burden with increasing time since the previous episode is progressively smaller as the post-treatment interval increases (Additional files 11, 12). In the most seasonal settings (MSI 90 %), these incremental values are smaller than in slightly less seasonal settings (MSI 70, 80 %). This is because the distribution of repeat malaria episodes peaks very sharply soon after a previous episode in extremely seasonal settings, and because malaria risk becomes very small when the post-episode window extends into the dry season. The advantage of LACTs with a longer mode of action than those currently available (e.g. an ACT that provided 10 weeks of PTP) may therefore be greater in settings where the distribution of repeat malaria episodes is spread slightly more evenly over time (i.e. high burden and seasonal, but not extremely seasonal). Other than where transmission is extremely seasonal, and particularly in perennial settings, longer PTP is always likely to be advantageous as although second events may occur close to the preceding episode in high transmission settings, subsequent episodes may also be prevented.
This analysis investigated the burden of repeated malaria episodes and the interval between successive episodes in cohort studies conducted in a range of epidemiological settings in West Africa, and used a model to estimate the effects of seasonality and transmission on the amount and timing of repeat malaria episodes across the continent. These results show that seasonality has an important effect on the distribution of recurrent malaria episodes, and should be considered alongside overall transmission intensity when deciding between different regimens for the case-management of malaria.
Repeated malaria attacks within a short time frame are not common where malaria transmission is low, and very few repeat episodes occur within the period of post-treatment prophylaxis provided by currently available drugs. This means that the additional benefit of a long-acting ACT over a shorter-acting drug in low transmission settings is likely to be minimal, and other factors such as cost, simplicity of dosing, tolerability and local resistance patterns to ACT partner drugs are likely to be more important considerations. Where malaria transmission is higher, repeated episodes of malaria are more common, and a larger fraction of the annual burden occurs within the period of post-treatment prophylaxis provided by longer-acting ACT, and therefore might be prevented if these regimens were used in place of shorter-acting regimens. In higher transmission settings, the burden is also larger in absolute terms, and the burden averted is also, therefore, likely to be larger.
At a given level of transmission intensity, the fraction of episodes occurring within a fixed time period after a previous attack increases in a non-linear fashion with increasing seasonality. This is plausible as in settings with a short season, any recurrent malaria must occur relatively close to a preceding episode, and because, to achieve a similar level of transmission in a shorter period, the transmission intensity must be higher during the season. Seasonality may also lead to more recurrent malaria because immunity is not continuously boosted as in perennial areas [37, 38] thus the development of the anti-disease immunity that reduces the probability of developing symptoms upon infection is delayed relative to perennial settings.
It is not certain how long protection against clinical malaria after LACT lasts, but a reasonable estimate might be between 4–8 weeks depending on the drug used, compared to 2–3 weeks for a shorter-acting ACT such as artemether-lumefantrine or artesunate-amodiaquine [12, 15, 39, 40]. These results show that the increase in the number of malaria episodes occurring between 28 and 42 days after a preceding attack was substantial in certain settings, suggesting that the additional prophylaxis will make important changes to the burden. This may be an underestimate if protection from a short-acting ACT lasts less than 28 days, or if protection from a LACT lasts longer than 6 weeks, or both. The decision to switch to ACT from non-ACT was primarily made on the basis of dramatic improvements in curative efficacy . However, when choosing between the artemisinin-based combinations that are now available, all of which have similarly high curative efficacy, more focus should be given to the potential benefits other than cure.
Since long-acting ACT appears to be of most value in areas of high and seasonal transmission, there is some overlap with seasonal malaria chemoprevention (SMC), which has recently been recommended in highly seasonal areas of the Sahel and sub-Sahel with high malaria burdens . SMC consists of monthly courses of sulfadoxine-pyrimethamine plus amodiaquine, and thus precludes use of two of the current ACTs: AS-SP or AS-AQ. However, the use of long-acting ACT for treatment, such as DHA-PQP or AS-MQ, may complement the SMC strategy, since a child with severe acute illness such as malaria will not receive a particular monthly course of SMC . A long-acting ACT for case management would help avoid the situation where experiencing malaria once predisposes a child to experiencing malaria again later in the transmission season by preventing them being protected by SMC.
This study has a number of limitations inherent to a model-based generalization of a more complex problem. Heterogeneity in exposure to malaria is likely to be influential, since this affects how unevenly malaria is shared between individuals, with the most exposed individuals in a particular setting most likely to experience recurrent malaria in a short period of time. Similarly, acquired immunity will affect the interval between malaria episodes. Acquisition of immunity may differ in highly seasonal settings due to waning of protection during long periods without exposure, and there being some redundancy in additional infections during the transmission season among individuals who are currently or have recently been infected; both heterogeneity and acquisition of immunity are captured by the model , but it is not easy to capture these patterns within data in order to validate these effects. However, the consistency of the patterns of repeat malaria seen in real data and the model estimates at different levels of transmission suggest that the modelling of heterogeneity and immunity do not pose a major problem in terms of the overall interpretation. The approach used in the present analysis focused on true malaria cases that presented for treatment, which in practice may not all be treated for malaria, but also ignores use of anti-malarials for non-malaria cases, which may provide some benefits. The Markham seasonality index used in this study can underestimate the level of seasonality in areas with bimodal patterns, but the approach used here, which identified and investigated bimodality separately, suggests that areas with a high degree of bimodality are not particularly common, and that bimodality simply increases the average interval between episodes depending on the relative intensity of the two peaks and the gap between the two rainy seasons.
A limitation in scope is that the model used does not include the impact of ACT in different settings on the development of drug resistance. Long-acting ACT may be more vulnerable to development of resistance because sub-therapeutic concentrations of anti-malarials may persist for long periods after treatment, although this is not the only factor that is important . Targeting long-acting ACT to areas where transmission is seasonal may help to reduce the rate at which resistance develops, by reducing the number of children who are exposed to reinfection while carrying sub-therapeutic concentrations of these drugs , and may help to prolong their useful life.
Areas of seasonal malaria transmission are likely to have more episodes of malaria that are preventable by switching to a long-acting drug for case management than areas of equivalent transmission spread more evenly over the year. This suggests that long-acting ACT should be considered carefully as an option for case-management in areas where malaria transmission is seasonal, particularly where malaria transmission is high.
MEC conceived and designed the study, analysed and interpreted the trial data, carried out the model-based simulations and wrote the first draft of the manuscript. PGTW, LCO, and TG assisted with model-based simulations and interpretation, and writing of the draft manuscript. JTG developed the individual-based malaria transmission model, and assisted with interpretation of the modelling results. KPA supervised field activities for the Kintampo cohort study and assisted with interpretation of the study data. SOA supervised field activities in the Kintampo Cohort Study and the Navrongo study, and assisted with interpretation of the trial data. DD supervised field activities for the Burkina IPTc study and assisted with interpretation of the trial data. AD supervised field activities for the Mali IPTc study and assisted with interpretation of the trial data. BC supervised field activities for the Niakhar IPTc study and assisted with interpretation of the trial data. DC supervised field activities in the Navrongo study and assisted with interpretation of the trial data. BMG contributed to study design and writing of the draft manuscript. ACG developed the individual-based malaria transmission model, contributed to study design, and assisted with interpretation of the modelling results. PJM contributed to study design, analysis of the data and writing of the draft manuscript. All authors contributed to interpretation of the analyses and revised the draft manuscript. All authors read and approved the final manuscript.
MC, PGTW and LCO are supported by Population Health Scientist Fellowships jointly funded by the UK Medical Research Council (MRC) and the UK Department for International Development (DFID) under the MRC/DFID Concordat agreement. JTG is supported by an MRC fellowship. ACG acknowledges grant funding from the Bill and Melinda Gates Foundation and Centre support from the MRC. The Navrongo IPTi Study was funded by the UK Department for International Development (DFID) (Grant no. R7602). The Kintampo Birth Cohort Study was funded by the US National Institutes of Health (Grant no: HHSN266200400016C). The Niakhar IPTc study received financial support from the Gates Malaria Partnership and the LSHTM DFID Malaria Knowledge Programme. The Burkina and Mali IPTc studies were supported by a grant to the London School of Hygiene & Tropical Medicine from the Bill & Melinda Gates Foundation (grant number: 41783).
Compliance with ethical guidelines
Competing interests The authors declare that they have no competing interests.
- 3.Taylor TE, Molyneux ME (2002) Clinical features of malaria in children. In: Warrell DA, Gilles HM (eds) Essential malariology. Arnold, LondonGoogle Scholar
- 7.Phiri K, Esan M, van Hensbroek MB, Khairallah C, Faragher B, ter Kuile FO (2012) Intermittent preventive therapy for malaria with monthly artemether-lumefantrine for the post-discharge management of severe anaemia in children aged 4–59 months in southern Malawi: a multicentre, randomised, placebo-controlled trial. Lancet Infect Dis 12:191–200CrossRefPubMedGoogle Scholar
- 13.Dorsey G: Modelling vs. real life settings: Longitudinal outcomes in a cohort of Ugandan children randomized to artemether-lumefantrine versus dihydroartemisinin-piperaquine for the treatment of malaria. In: American Society of Tropical Medicine and Hygiene. Atlanta, Georgia 2012Google Scholar
- 18.Konate AT, Yaro JB, Ouedraogo AZ, Diarra A, Gansane A, Soulama I et al (2011) Intermittent preventive treatment of malaria provides substantial protection against malaria in children already protected by an insecticide-treated bednet in Burkina Faso: a randomised, double-blind, placebo-controlled trial. PLoS Med 8:e1000408PubMedCentralCrossRefPubMedGoogle Scholar
- 21.Dicko A, Diallo AI, Tembine I, Dicko Y, Dara N, Sidibe Y et al (2011) Intermittent preventive treatment of malaria provides substantial protection against malaria in children already protected by an insecticide-treated bednet in Mali: a randomised, double-blind, placebo-controlled trial. PLoS Med 8:e1000407PubMedCentralCrossRefPubMedGoogle Scholar
- 22.Cisse B, Sokhna C, Boulanger D, Milet J, Ba EH, Richardson K et al (2006) Seasonal intermittent preventive treatment with artesunate and sulfadoxine-pyrimethamine for prevention of malaria in Senegalese children: a randomised, placebo-controlled, double-blind trial. Lancet 367:659–667CrossRefPubMedGoogle Scholar
- 41.WHO (2006) Guidelines for the treatment of malaria. World Health Organization, GenevaGoogle Scholar
- 42.WHO policy recommendation: Seasonal malaria chemoprevention (SMC) for Plasmodium falciparum malaria control in highly seasonal transmission areas of the Sahel sub-region in Africa (http://www.who.int/malaria/publications/atoz/who_smc_policy_recommendation/en/)
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.