**4. Contribution of low-density asymptomatic infections to transmission**

In the previous sections, we discussed factors influencing the duration and average density of individual infections. In this section, our goal is to understand the significance and contribution of low-density asymptomatic infections to local transmission. This question is particularly important in areas where control efforts have pushed transmission towards near elimination levels – in this case it has been hypothesised that chronic low density asymptomatic infections could maintain local transmission.

from Vietnam suggest that chronic sub-patent infections can lead to high parasitaemias, 5–6

**Figure 2.** Estimated infectiousness to mosquitoes over course of infection and cumulative relative contribution to onwards infectiousness for each type of infection. The infectiousness of each parasite density trajectory from **Figure 1** is estimated by assuming that individuals with very high parasite densities (associated with being febrile) are three times as infectious as individuals with microscopy-detectable asymptomatic infection, who are then in turn three times as infectious as individuals with sub-microscopic asymptomatic infection [32, 56]. The cumulative infectivity of an individual is simply the area under the infectiousness curve (left panel). This area under the curve of each type of infection is compared in

Understanding the Importance of Asymptomatic and Low-Density Infections…

http://dx.doi.org/10.5772/intechopen.77293

137

The relative proportions of low density infections which go on to rapidly clear *versus* those which become higher density infections are unknown and likely depends on many of the factors highlighted in Sections 2 and 3, such as age, immunity and host genetics. Studying variations in detectability over the course of a single naturally acquired infection is difficult for a few reasons: (i) super- and co-infections: in high transmission settings, most individuals are infected with more than one parasite clone [54] and standard techniques do not indicate the density of each parasite genotype (therefore the density of older *versus* newer infections cannot be distinguished); (ii) even using molecular methods, parasite densities often fluctuate below detection limits before the end of an infection and it is difficult to distinguish this from clearance of infection; (iii) long follow up is needed: even in endemic areas where individuals have immunity, specific parasite genotypes have been shown to persist for more than 6 months [55]. Here, we use a simple modelling framework to explore how the duration and

infectiousness of an infection affect the impact of different intervention strategies.

Four archetypal parasite density trajectories are identified from the malaria therapy data [57, 58] to represent broadly four potential outcomes of a new infection (**Figure 3**): (1) initially symptomatic before becoming asymptomatic and fluctuating between patent and sub-patent levels for a long time (~300 days); (2) initially symptomatic before becoming asymptomatic and fluctuating

orders of magnitude higher [11].

the right panel.

**4.2. Model framework**

#### **4.1. Should we detect and treat low-density asymptomatic infections?**

Since identifying the reservoir of low-density infections, there has been interest in developing more sensitive rapid diagnostics in order to detect and treat these infections. However, the benefit of treating such infections, both at the individual level and in terms of preventing onward transmission to others, remains unclear. The impact of treating a low-density infection depends not only on its current infectiousness to mosquitoes, but the future course of infection and infectiousness that is prevented (**Figures 1** and **2**). If low-density infections most commonly represent the tail end of an infection, which will clear rapidly without treatment, then the benefit of treatment would be small (**Figure 2**, yellow bar). However, if such infections commonly become chronic and lead to future periods of higher parasite densities, infectiousness, and possibly also symptoms, the benefit of treating such infections would be greater (**Figure 2**, green and light blue bars). Consistent with this second scenario, longitudinal data

#### Understanding the Importance of Asymptomatic and Low-Density Infections… http://dx.doi.org/10.5772/intechopen.77293 137

**Figure 2.** Estimated infectiousness to mosquitoes over course of infection and cumulative relative contribution to onwards infectiousness for each type of infection. The infectiousness of each parasite density trajectory from **Figure 1** is estimated by assuming that individuals with very high parasite densities (associated with being febrile) are three times as infectious as individuals with microscopy-detectable asymptomatic infection, who are then in turn three times as infectious as individuals with sub-microscopic asymptomatic infection [32, 56]. The cumulative infectivity of an individual is simply the area under the infectiousness curve (left panel). This area under the curve of each type of infection is compared in the right panel.

from Vietnam suggest that chronic sub-patent infections can lead to high parasitaemias, 5–6 orders of magnitude higher [11].

The relative proportions of low density infections which go on to rapidly clear *versus* those which become higher density infections are unknown and likely depends on many of the factors highlighted in Sections 2 and 3, such as age, immunity and host genetics. Studying variations in detectability over the course of a single naturally acquired infection is difficult for a few reasons: (i) super- and co-infections: in high transmission settings, most individuals are infected with more than one parasite clone [54] and standard techniques do not indicate the density of each parasite genotype (therefore the density of older *versus* newer infections cannot be distinguished); (ii) even using molecular methods, parasite densities often fluctuate below detection limits before the end of an infection and it is difficult to distinguish this from clearance of infection; (iii) long follow up is needed: even in endemic areas where individuals have immunity, specific parasite genotypes have been shown to persist for more than 6 months [55]. Here, we use a simple modelling framework to explore how the duration and infectiousness of an infection affect the impact of different intervention strategies.

#### **4.2. Model framework**

this problem and less than 30% of all falciparum infections are estimated to be missed by this method [48]. The timing of blood sampling in parasitological surveys might also affect parasite detection and quantification because asexual falciparum parasites do not circulate continuously; sequestration of falciparum schizonts starts 12-18 hours after merozoite invasion and during this period they might not be detectable. An intensive longitudinal study in Tanzania showed that periodic changes in parasite densities are common. The periodicity of clone-specific detectability indicates that in natural infections, synchronised sequestration of clonal parasite populations occurs [49]. A study [50] that collected samples on two consecutive days found a prevalence disparity of approximately 25% between the two samples. Periodic changes in parasite levels could have a direct impact on the selection of diagnostics,

for example by favouring assays that detect more persistent markers, such as HRP-2.

**4. Contribution of low-density asymptomatic infections to** 

**4.1. Should we detect and treat low-density asymptomatic infections?**

atic infections could maintain local transmission.

infectivity.

**transmission**

136 Towards Malaria Elimination - A Leap Forward

The detection of either asexual or sexual stage parasites is sufficient to establish the diagnosis of infection. Although gametocytes are not known to periodically sequester, there is evidence of periodic variation in gametocyte levels [51] in peripheral blood. For several decades now [52], accumulation of mature gametocytes in the skin [53] has been hypothesised as a possible mechanism of transmission enhancement. If confirmed, this would imply that subpatent gametocytaemias in peripheral blood might be associated with higher-than-expected

In the previous sections, we discussed factors influencing the duration and average density of individual infections. In this section, our goal is to understand the significance and contribution of low-density asymptomatic infections to local transmission. This question is particularly important in areas where control efforts have pushed transmission towards near elimination levels – in this case it has been hypothesised that chronic low density asymptom-

Since identifying the reservoir of low-density infections, there has been interest in developing more sensitive rapid diagnostics in order to detect and treat these infections. However, the benefit of treating such infections, both at the individual level and in terms of preventing onward transmission to others, remains unclear. The impact of treating a low-density infection depends not only on its current infectiousness to mosquitoes, but the future course of infection and infectiousness that is prevented (**Figures 1** and **2**). If low-density infections most commonly represent the tail end of an infection, which will clear rapidly without treatment, then the benefit of treatment would be small (**Figure 2**, yellow bar). However, if such infections commonly become chronic and lead to future periods of higher parasite densities, infectiousness, and possibly also symptoms, the benefit of treating such infections would be greater (**Figure 2**, green and light blue bars). Consistent with this second scenario, longitudinal data

Four archetypal parasite density trajectories are identified from the malaria therapy data [57, 58] to represent broadly four potential outcomes of a new infection (**Figure 3**): (1) initially symptomatic before becoming asymptomatic and fluctuating between patent and sub-patent levels for a long time (~300 days); (2) initially symptomatic before becoming asymptomatic and fluctuating

last 14 days) at any time and will start at the beginning of a new parasite density trajectory selected based on their age-specific probability of developing symptoms. The impact of two interventions is simulated: increasing treatment coverage among febrile individuals to 90% or delivering a single round of MDA to a random 80% of the population. The effect of each intervention is assessed by calculating the percentage reduction in the combined onwards infectiousness of the whole population in the following year, which depends on whether they are in a patent and symptomatic, patent and asymptomatic or sub-patent state. Parasite density is translated to infectiousness according to assumptions described in **Figure 2**. After an intervention the infection risk is reduced proportionally with the reduction in the proportion of the population that are infected to account for the population-level impact of these inter-

High transmission (20% slide prevalence) 66% 52% 38% Low transmission (5% slide prevalence) 78% 70% 59%

**Table 1.** Probability of developing symptoms upon being successfully inoculated (estimates from [56]).

**Age range (in years) 0–5 5–15 15+**

http://dx.doi.org/10.5772/intechopen.77293

139

Understanding the Importance of Asymptomatic and Low-Density Infections…

MDA is predicted to be more effective at reducing the infectious reservoir than increasing treatment coverage among febrile individuals in low transmission settings with both short and long infection durations and high transmission settings with long infection durations only (**Figure 4**). In these scenarios the rebound of infection is slow, meaning over the course of a year, MDA prevents a higher number of infected/infectious days than increasing treatment coverage of febrile individuals. In high transmission settings with a short duration of infection, a higher force of infection is needed to achieve a given prevalence. Therefore, the effect of any intervention is reduced because the population become reinfected quicker. In this scenario, increasing treatment coverage is more effective because it is a sustained intervention. It is important to note that the outcome metric considered here is the reduction in the infectious reservoir increasing treatment coverage is likely to always have the greatest impact on reducing malaria morbidity and mortality in all transmission scenarios. The model simply illustrates how our uncertainties about the duration of untreated infection affect estimates of intervention impact.

Since asymptomatically infected individuals do not actively seek antimalarial treatment, their infections may last longer than symptomatic episodes. In this chapter, we discussed human and parasite factors that influence the dynamics of parasitaemia and the duration of gametocyte circulation in these infections. These factors result in a range of infection profiles, the relative combinations of which in a population will define not only the composition of the infectious reservoir but the likelihood of success of intervention measures. For example, our calculations suggest that MDA is most effective if infections have long durations. In a high transmission setting, MDA might have been expected to be more effective than increasing

ventions on transmission.

**5. Conclusions**

**Figure 3.** Parasite density trajectories from four infected individuals. The arrows represent the estimated time to clear infection (assuming a period of sub-patent infection after the last patent day of infection). The upper panel shows two symptomatic patients (red points indicate the days on which the patients were febrile) and the lower panel shows asymptomatic patients. The horizontal dotted line indicates the limit of detection of field microscopy (100 parasites/μl).

between patent and sub-patent levels for a short time (~120 days); (3) always asymptomatic and fluctuating between patent and sub-patent densities for a long time (~300 days); and (4) always asymptomatic and fluctuating between patent and sub-patent for a short time (~90 days). Note how these relate to the hypothesised profiles in **Figure 1**.

An age-structured population of individuals is simulated whereby individuals have a daily probability of acquiring a new infection. Upon being infected, an individual's probability of developing symptoms is based on their age and the intensity of transmission (fitted estimates taken from [56]) (**Table 1**). In a single simulation, infections are assumed to be either all long (300 days) or all short (120 or 90 days). Infected individuals will follow one of the parasite density trajectories shown in **Figure 3** unless they are treated or re-infected. Febrile individuals have a 50% probability of receiving treatment. Treated individuals are assumed to clear their asexual parasites after being febrile for 3 days, they then become non-infectious after 6 days. These individuals are also assumed to be protected from reinfection for 14 days after treatment.

The model is simulated with either high transmission (20% slide prevalence) or low transmission (5% slide prevalence) and the daily probability of infection is fitted to achieve these prevalence levels. Infected individuals can be reinfected (unless they received treatment in the


**Table 1.** Probability of developing symptoms upon being successfully inoculated (estimates from [56]).

last 14 days) at any time and will start at the beginning of a new parasite density trajectory selected based on their age-specific probability of developing symptoms. The impact of two interventions is simulated: increasing treatment coverage among febrile individuals to 90% or delivering a single round of MDA to a random 80% of the population. The effect of each intervention is assessed by calculating the percentage reduction in the combined onwards infectiousness of the whole population in the following year, which depends on whether they are in a patent and symptomatic, patent and asymptomatic or sub-patent state. Parasite density is translated to infectiousness according to assumptions described in **Figure 2**. After an intervention the infection risk is reduced proportionally with the reduction in the proportion of the population that are infected to account for the population-level impact of these interventions on transmission.

MDA is predicted to be more effective at reducing the infectious reservoir than increasing treatment coverage among febrile individuals in low transmission settings with both short and long infection durations and high transmission settings with long infection durations only (**Figure 4**). In these scenarios the rebound of infection is slow, meaning over the course of a year, MDA prevents a higher number of infected/infectious days than increasing treatment coverage of febrile individuals. In high transmission settings with a short duration of infection, a higher force of infection is needed to achieve a given prevalence. Therefore, the effect of any intervention is reduced because the population become reinfected quicker. In this scenario, increasing treatment coverage is more effective because it is a sustained intervention. It is important to note that the outcome metric considered here is the reduction in the infectious reservoir increasing treatment coverage is likely to always have the greatest impact on reducing malaria morbidity and mortality in all transmission scenarios. The model simply illustrates how our uncertainties about the duration of untreated infection affect estimates of intervention impact.

#### **5. Conclusions**

between patent and sub-patent levels for a short time (~120 days); (3) always asymptomatic and fluctuating between patent and sub-patent densities for a long time (~300 days); and (4) always asymptomatic and fluctuating between patent and sub-patent for a short time (~90 days). Note

**Figure 3.** Parasite density trajectories from four infected individuals. The arrows represent the estimated time to clear infection (assuming a period of sub-patent infection after the last patent day of infection). The upper panel shows two symptomatic patients (red points indicate the days on which the patients were febrile) and the lower panel shows asymptomatic patients. The horizontal dotted line indicates the limit of detection of field microscopy (100 parasites/μl).

An age-structured population of individuals is simulated whereby individuals have a daily probability of acquiring a new infection. Upon being infected, an individual's probability of developing symptoms is based on their age and the intensity of transmission (fitted estimates taken from [56]) (**Table 1**). In a single simulation, infections are assumed to be either all long (300 days) or all short (120 or 90 days). Infected individuals will follow one of the parasite density trajectories shown in **Figure 3** unless they are treated or re-infected. Febrile individuals have a 50% probability of receiving treatment. Treated individuals are assumed to clear their asexual parasites after being febrile for 3 days, they then become non-infectious after 6 days. These indi-

viduals are also assumed to be protected from reinfection for 14 days after treatment.

The model is simulated with either high transmission (20% slide prevalence) or low transmission (5% slide prevalence) and the daily probability of infection is fitted to achieve these prevalence levels. Infected individuals can be reinfected (unless they received treatment in the

how these relate to the hypothesised profiles in **Figure 1**.

138 Towards Malaria Elimination - A Leap Forward

Since asymptomatically infected individuals do not actively seek antimalarial treatment, their infections may last longer than symptomatic episodes. In this chapter, we discussed human and parasite factors that influence the dynamics of parasitaemia and the duration of gametocyte circulation in these infections. These factors result in a range of infection profiles, the relative combinations of which in a population will define not only the composition of the infectious reservoir but the likelihood of success of intervention measures. For example, our calculations suggest that MDA is most effective if infections have long durations. In a high transmission setting, MDA might have been expected to be more effective than increasing

absence of repeated rounds of MDA, increasing treatment coverage may in fact be more effective in the long term (**Figure 4**). However, the model assumes no seasonality, when in reality many malaria endemic regions transmission is highly seasonal; this could underestimate the impact of MDA. As discussed above, seasonal changes in infection duration represent another aspect of the epidemiology of asymptomatic infections that could be explored to target interventions. Indeed, where transmission is seasonal, infections persisting during the dry season correspond to long-term asymptomatic carriage since incidence of new infections is thought to be negligible. This means that during this period, infections are likely to be missed by pas-

Understanding the Importance of Asymptomatic and Low-Density Infections…

http://dx.doi.org/10.5772/intechopen.77293

141

sive surveillance, while active approaches, such as MDA, might be more efficacious.

determine if and what additional measures are required for malaria elimination.

CD and BG would like to acknowledge funding from the Bill & Melinda Gates Foundation (OPP1034789 & OPP1173572). LO is funded by a UK Royal Society Dorothy Hodgkin Fellowship, and also acknowledges grants from the Bill & Melinda Gates Foundation and Medicines for Malaria Venture. HS is supported by an Imperial College junior research

**Acknowledgements**

**List of abbreviations**

HbC haemoglobin C HbS haemoglobin S

HbAA haemoglobin A (homozygous) HbAS haemoglobin AS (heterozygous)

fellowship.

Determining the optimal control strategy, and moreover, whether asymptomatic/sub-patent infections actually need to be identified and treated, will require careful analysis of local epidemiological data. The three key metrics that need to be determined are: (1) the proportion of individuals that develop symptoms and seek treatment, (2) the distribution of durations of asymptomatic infections, and (3) the relative infectivity of different infections. These factors are in turn driven by the complex interplay of host immunological factors, such as strain-specific immunity, intrinsic parasite growth factors and population characteristics (e.g. prevalence of HbAA *versus* HbAS, variation in demographic risk within a community). The relative high prevalence of asymptomatic and low-density infections in areas with low transmission and high treatment coverage might indicate that either these infections are contributing towards transmission and enabling malaria to persist or that they reflect the tail end of infections with transmission maintained by the few highly infectious symptomatic cases. This will vary in different settings and whilst the rapid identification and treatment of symptomatic malaria infections remains key to all control approaches, a better understanding of the nature of asymptomatic infections will

**Figure 4.** Simulated impact of increasing treatment coverage or mass drug administration (MDA) on the proportion of mosquitoes infected by a population: influence of transmission setting and infection duration. The grey lines represent continuing 50% treatment coverage and no MDA, and the blue and red lines as shown in the legend.

treatment coverage, because higher immunity reduces the probability of developing symptoms and the proportion of infections getting treated. However, when transmission is high, the reduction in prevalence after an MDA is temporary, due to the drug half-life and imperfect coverage levels, and individuals are likely to become reinfected quickly, therefore in the absence of repeated rounds of MDA, increasing treatment coverage may in fact be more effective in the long term (**Figure 4**). However, the model assumes no seasonality, when in reality many malaria endemic regions transmission is highly seasonal; this could underestimate the impact of MDA. As discussed above, seasonal changes in infection duration represent another aspect of the epidemiology of asymptomatic infections that could be explored to target interventions. Indeed, where transmission is seasonal, infections persisting during the dry season correspond to long-term asymptomatic carriage since incidence of new infections is thought to be negligible. This means that during this period, infections are likely to be missed by passive surveillance, while active approaches, such as MDA, might be more efficacious.

Determining the optimal control strategy, and moreover, whether asymptomatic/sub-patent infections actually need to be identified and treated, will require careful analysis of local epidemiological data. The three key metrics that need to be determined are: (1) the proportion of individuals that develop symptoms and seek treatment, (2) the distribution of durations of asymptomatic infections, and (3) the relative infectivity of different infections. These factors are in turn driven by the complex interplay of host immunological factors, such as strain-specific immunity, intrinsic parasite growth factors and population characteristics (e.g. prevalence of HbAA *versus* HbAS, variation in demographic risk within a community). The relative high prevalence of asymptomatic and low-density infections in areas with low transmission and high treatment coverage might indicate that either these infections are contributing towards transmission and enabling malaria to persist or that they reflect the tail end of infections with transmission maintained by the few highly infectious symptomatic cases. This will vary in different settings and whilst the rapid identification and treatment of symptomatic malaria infections remains key to all control approaches, a better understanding of the nature of asymptomatic infections will determine if and what additional measures are required for malaria elimination.
