In two cohorts from two geographically distinct environments, S. japonicum infections repeatedly occurred in the same individuals over time, following treatment with praziquantel. This clustering of infections occurred even when accounting for exposure, and clustering was particularly strong in cohort 2, a population with low overall infection prevalence and intensity. These findings suggest there exists a subset of individuals within the general population that is particularly vulnerable to S. japonicum infection. Alternatively, this subset of individuals may have uncured infections due to non-compliance or treatment failure. This has important implications for disease surveillance: individuals with a history of S. japonicum infection may serve as appropriate targets for infection monitoring and treatment in low-prevalence environments. In addition, our findings provide evidence for host susceptibility to helminth infections – suggesting some individuals may be more vulnerable to infection given equivalent exposures.
It is possible that individuals who are repeatedly infected with helminthes are simply the most highly exposed individuals in the population. Cercarial exposure is a well-documented determinant of S. japonicum
. We found that the ratio of observed to expected prevalence of consecutive infections exceeded unity using an exposure-blind prediction model. This ratio was lower when we included exposure in the prediction models, but still exceeded unity. This suggests some individuals may be repeatedly infected due to their high cercarial exposure, but exposure does not fully explain this phenomenon. S. japonicum
exposure is challenging to assess due to the difficulties in quantifying daily human behaviors and the absence of practical methods for directly measuring cercarial concentration, and our prediction models are limited by our ability to accurately measure cercarial exposure. However, the imperfections of our exposure measures are likely offset by the use of an aggressive, data adaptive algorithm to predict S. japonicum
infection using over 25 exposure variables. Over-fitting is possible when using such methods, which, in this case, would have a conservative impact on our estimates, pushing observed to expected ratios closer to unity. Therefore, exposure alone is unlikely to explain the observed concentration of repeated schistosomiasis infections in a subset of the population.
More likely, individuals who are repeatedly infected with helminthes may be those who have a sufficiently elevated combination of susceptibility and exposure. We explored the clustering of infections within certain individuals from a mechanistic perspective by postulating that an individual's worm burden, w
, accumulated from exposures subsequent to successful praziquantel treatment, can be described at the end of one or more infection seasons as a result of that individual's cumulative exposure to cercariae, E
, and the subsequent penetration and development of a fraction of these cercarial hits, α
, into adult parasites. That is, w=αE
is composed of two elements, water contact, S
, and cercarial concentration, C.
The parameter α
, reflecting host susceptibility, is assumed to be a stable property of each individual in the village population and the distribution of C is assumed to be a village property shared by all inhabitants. The water contact measurements described above and cercarial bioassay data collected in conjunction with the prior studies of cohort 1 
suggest that the population distribution of E
is strongly right skewed as is generally observed to be the case for distributions of w
Assuming that the distributions of exposure and susceptibility in a population are independent, their joint distribution is depicted in . The marginal distribution of exposures, f(E)
, is for illustrative purposes shown as a negative exponential distribution since multiple cercarial hits are thought to be necessary to lead to a single adult worm. Also for illustration, the marginal distribution of susceptibility, h(α)
, is shown as symmetric. The line wT=αE
is the threshold of infections that are epidemiologically visible which we define here as the minimum worm burden necessary to produce eggs at the lower limit of detection by a combination of the miracidia hatching test and the Kato-Katz method. The fraction of the population susceptible to infection at or above this threshold is that lying to the right of the line α
*. That is, the probability of an exposure leading to a diagnosis of infection for an individual with an α
less than α
* is essentially zero given the maximum cercarial exposure in this hypothetical environment. The shaded area depicts the set of exposure-susceptibility combinations that produce detectable infections.
The marginal distributions of exposure and host susceptibility, together with contours of their joint distribution.
Specification of the two marginal distributions allows the calculation of the distribution of their product, that is, the distribution of worm burden in the population. However, the point here is that, at least in this generic example, the proportion of the population at risk for infection is less than the entire population. That is, the number of individuals susceptible to infection,
, in this environment is:
is the total population size. Hence, if
is the observed number of infections, the ratio of prevalence of infection in the susceptible population to the total population is:
which is always equal to or greater than unity.
Returning to the re-infection issue, suppose the population is exposed in an unchanging environment, treated annually with praziquantel at T
1, and T
2, and infection assessed at the end of year 1 and year 2. Since the same population is at risk of infection with the same marginal distribution of exposure in both years, and this population is less than the entire population, the observed number of repeated infections will be greater than that expected based on infections occurring randomly in the entire population. It follows that the ratio of observed re-infections to the expected number, if distributed randomly in the entire population, is simply the square of the foregoing equation:
Moreover, as the fraction of exposure-susceptibility combinations that produce infection decreases, α*
and this ratio both increase. Hence, the simple model of the infection process with individual differences in susceptibility to infection, depicted in ,provides a heuristic explanation of the epidemiological finding that the ratio of observed to expected re-infections increases as prevalence of infection decreases. Clearly, more refined analyses are possible that address a more rigorous definition of α*
, take distributional assumptions into account, or explore the effect of variability in individual water contact. We will further address these and related determinants of transmission in the low-risk environment via an individually-based stochastic model which will be the subject of a future report. In addition, it is possible to estimate the proportion of susceptibles in a population via a statistical innovation using a model selection procedure like SuperLearner in the context of a latent mixture model, where the susceptibility status is latent – an approach that we will pursue in the future.
The factors that govern α
are not fully characterized for schistosomiasis or other helminthiases. However, there is substantial evidence that immune function, particularly the ability to mount antigen-specific IgE response, can confer host resistance to schistosomiasis as well as other helminthiases 
. Immune response is likely attributable to a combination of past exposure, treatment and host genetics 
. Physical characteristics such as skin thickness may also play a role in determining host resistance or susceptibility. As these genetic and immunological pathways are further elucidated, the definition of α
may be further refined.
Alternatively, it is possible that the individuals who appear to be repeatedly infected with S. japonicum
do not have new infections, but instead have residual, uncured infections that persist despite treatment. Praziquantel is the primary drug used to treat schistosomiasis infections, and resistance is an ongoing concern, particularly in areas where the drug has been used extensively. In China, praziquantel has been widely administered since the 1990s through mass and targeted treatment campaigns. Currently, there is no evidence of population-level resistance to S. japonicum
, S. haematobium
or S. mansoni
, but praziquantel resistant laboratory isolates have been identified 
. It is possible that praziquantel kills some but not all parasites, resulting in an incomplete cure. Repeated dosing with praziquantel may enhance treatment efficacy, particularly for individuals with high infection intensities 
. While infection intensities in our two cohorts were generally low, we cannot rule out the possibility that what appear to be repeated infections are, in fact, infections that were not cured by praziquantel treatment.
Uncured S. japonicum
infection may also be the result of poor adherence to drug treatment. As schistosomiasis morbidity declines, it is possible that so too do the perceived risks of infection and willingness to take praziquantel. Praziquantel has an excellent safety record and is appropriate for mass drug distribution, even in very young populations 
but the drug has a bitter taste and can cause transient side effects, including nausea and dizziness. In a recent survey, 33% of people said such side effects impacted their ability to work 
. We have found a high degree of self-reported treatment adherence (>90%) in surveys of 236 people drawn from the same villages as cohort 1 (surveyed in 2007) and 686 people drawn from the same villages as cohort 2 (surveyed in 2008), but other studies have documented poor compliance with mass-treatment campaigns for helminthiasis 
. Our findings underscore the importance of continued monitoring of treatment effectiveness, including both drug resistance and population perceptions of the risks and benefits of treatment. Methods
capable of distinguishing new from residual infections could advance our understanding of treatment efficacy and drug adherence.
Our findings underscore surveillance challenges in areas where worm burdens are low. While individuals with high worm burdens have the potential to contribute a large number of future infections, our prior work suggests that even modest parasite inputs are sufficient to sustain schistosomiasis transmission 
. In China, surveillance and elimination efforts are made more complex as there are at least 40 competent mammalian host species for S. japonicum
, and bovines are suspected to be key reservoirs in some areas 
. Thus the ability to identify humans and, in the case of S. japonicum
, other mammalian hosts with low-intensity helminth infections may be crucial to efforts to prevent the reemergence of helminth infections in areas where disease control efforts have successfully lowered infections and morbidity. Many of the individuals who tested positive for S. japonicum
in our study had worm burdens below the limit of detection of the Kato-Katz assay, the schistosomiasis diagnostic method recommended by the World Health Organization 
. Immunoassays generally have high sensitivity, but it can be difficult to distinguish past from current infections, which is particularly problematic when attempting to identify residual infections in regions with previously high infection prevalence and intensity 
. While new methods offer promise 
, the current lack of practical, highly sensitive diagnostics is a barrier to the long-term control of helminthiases 
As China aims to eliminate schistosomiasis and global efforts are launched to eliminate a number of helminthiases, the success of such efforts may hinge, in part, on the ability to identify reservoirs of infection and reduce the potential of such reservoirs to generate future infections. Our findings suggest that there exist an identifiable, high-risk subpopulation for S. japonicum
infection. Due to high exposure, host susceptibility or treatment failure, these individuals are potential future reservoirs of S. japonicum
. Further, as infection prevalence declines, and with it, cercarial exposure, we expect the fraction of the population that is susceptible to S. japonicum
infection to decline. Thus, as regions approach disease control goals, targeted interventions may prove efficient and effective. In low-prevalence regions, individuals who test positive for S. japonicum
should be tested regularly and provided pharmaceutical treatment and transmission-blocking interventions such as improved household latrines