Background
The global burden of malaria has decreased markedly in the last decades due in large parts to increased funding, widespread deployment of insecticide-treated bed-nets, better diagnostics and enhanced availability of artemisinin-based combination therapies [
1‐
3]. Since the turn of the century, malaria incidence worldwide has fallen by more than 60% and 17 countries have eliminated malaria completely, fueling optimism of global malaria eradication [
3‐
7]. There is a growing awareness, however, that eradicating malaria will require a much better understanding of the
Plasmodium vivax parasite and how to successfully address the challenges it presents. Unfortunately, research on
P. vivax malaria has lagged far behind research on
Plasmodium falciparum malaria, and there are currently critical gaps in knowledge across a broad range of topics on
P. vivax [
8‐
14]. With accumulating evidence that
P. vivax infection is less benign than once thought [
9,
11,
12,
15,
16] and that
P. vivax prevalence might be underestimated in regions where
P. vivax and
P. falciparum coexist [
17‐
19], there is a clear need for further studies.
Even though
P. vivax malaria primarily affects countries in South and Southeast Asia,
P. vivax malaria remains an important public health problem around the Horn of Africa [
8,
10]. In Ethiopia, more than 50 million people are estimated to live in malaria-risk zones causing between 5 and 10 million clinical infections of malaria per year, and the rate of
P. vivax infection is comparable to that of
P. falciparum [
20‐
23]. The number of studies on
P. vivax malaria in the country has been steadily increasing. However, because vivax malaria is characterized by a high degree of local variation—in part caused by marked regional differences in altitude, temperature and rainfall—knowledge gained from one setting cannot necessarily be transferred to another [
24,
25]. Moreover, malaria control programmes have been scaled up markedly in the last decade, meaning results obtained from studies in the early parts of the century or before might be outdated due to changed transmission intensities and dynamics [
26,
27].
Various risk factors for
P. vivax infection have been investigated around the Horn of Africa, including climatic parameters, bed net use, indoor residual spraying, household characteristics and demographic factors [
23,
28‐
41]. A few studies have investigated household proximity to vector breeding site as a risk factor, though the evidence from these studies is mixed: three studies found an enhanced risk [
29,
30,
33], one found no difference [
28], and one found a reduction in risk with increasing proximity [
40]. A separate study concluded that the association between proximity to breeding site and
P. vivax infection is modified by age, with proximity a much stronger risk factor in children compared to adults [
38]. Further research on the topic is warranted.
The aim of the present study was to investigate the association between distance to vector breeding site and P. vivax infection in rural Ethiopia, and, secondarily, to examine if age was an effect modifier in the association.
Discussion
This 2-year prospective cohort study conducted between 2009 and 2011 in the rural Chano Mille Kebele in southern Ethiopia examined the association between P. vivax infection and household distance to vector breeding site using Cox regression modelling. The study found strong evidence of increasing rates of P. vivax infection the closer participants lived to the breeding site after adjusting for age, gender, persons living in household, education head of household, household wealth and ITN use. The final regression model showed that the group of participants who lived closest to the breeding site had almost 3.5 times the rate of P. vivax infection compared to the group who lived furthest away. The effect of increasing proximity to breeding site on P. vivax infection rate appeared to be linear throughout the distance categories investigated. There was no evidence that age was an effect modifier, investigated as a secondary objective in the study. Other important, though post hoc, findings included strong evidence that age was negatively associated with P. vivax infection (rates were more than 5 times higher in the age group 0–4 years compared to adults 25 years or older), and that overall ITN use was associated with reduced P. vivax infection rates. There was also some indication that the increased rate of P. vivax infection associated with younger age was more pronounced in the time-period when overall reported ITN use was low in the population (i.e. before the net distribution campaign).
The main finding of the present study, that there was a clear positive association between household proximity to vector breeding site and
P. vivax infection, is in line with the study’s pre-specified hypothesis and consistent with the majority of evidence from similar research conducted around the Horn of Africa [
29,
30,
33,
38]. Loha et al. also found a positive association between proximity to breeding site and
Plasmodium falciparum infection in the same study population [
42]. Given how malaria is transmitted, it is biologically plausible that individuals in households close to where mosquitoes live and breed have increased risk of infection [
16]. The female anopheline mosquito carrying the malaria parasite requires a blood-meal after mating in order for eggs to develop, and as the female leaves the breeding site in search of a suitable human host, it is logical that individuals who are closer to where the search starts have a higher likelihood of being targeted. However, it is not necessarily household proximity to breeding site per se that matters, rather, household proximity might be considered a strong positive correlate of the relative probability of being bitten and infected. This probability also depends on many other factors. For example, Loha et al. found evidence that the risk of malaria infection for individuals in a given household at a given distance to the breeding site was reduced if there were many other households lying between their household and the breeding site [
36]. In other words, the spatial layout of households in a village in relation to the breeding site might influence malaria infection risk and modify the effect of proximity in and of itself. Similarly, it is possible that other physical characteristics of a village or region—e.g. topography or vegetation—might modify the effect of proximity to breeding site.
The strength of the findings is supported by the clear stepwise increase in the rate of
P. vivax infection with increasing proximity to breeding site. The evidence of linearity between proximity and
P. vivax infection is somewhat at odds with findings from another study in Ethiopia where the relationship tended be exponential [
34]. Importantly though, the distance range investigated in that study was 150–1250 m compared to 1646–3717 m in the present study. It is possible that
P. vivax infection rates increase exponentially with proximity to breeding site at distances close to the breeding site and more linearly at distances further away.
As discussed in the limitation section below, it is hard to exclude selection bias resulting from loss to follow-up. However, when the follow-up period was split at the point of the net distribution campaign, the evidence of an association between proximity to breeding site and P. vivax infection was strong for both periods even though the population in the two periods differed (due to both follow-up loss and the entry of new participants). This makes it less likely that selection bias has greatly distorted the main conclusion. Misclassification and residual confounding can also not be excluded.
In terms of the study’s secondary objective of investigating age as an effect modifier, there was no evidence neither in bivariable nor multivariable analysis to support this. This finding contrasts an earlier study by Peterson et al. which found the effect of proximity to breeding site to be more pronounced in younger age groups [
38]. Several reasons might explain this difference. One, the present study was conducted in a rural setting at an altitude of about 1200 m versus the study by Peterson et al. which was done in a peri-urban setting at an altitude of 1600 m. Another important difference is that the study by Peterson et al. was of shorter duration and conducted during peak malaria season with consequently higher incidence rates of malaria over the course of the study period (the present study covered both peak and non-peak season). Though perhaps the most important differences between the studies were that the overall ITN use fraction was much lower in the study by Peterson et al. and that the distance-range investigated was much closer to the breeding site.
The post hoc finding that younger age was strongly associated with higher rates of
P. vivax infection is in line with prior research [
30,
37,
39], and in accordance with current understanding of the development of
P. vivax immunity [
49]. Since immunity against
P. vivax develops at an earlier age than immunity against
P. falciparum, is was unsurprising that the youngest age group had significantly higher
P. vivax rates compared to the second youngest group in the present study, however, that
P. falciparum rates in the same population were roughly equal in the two age groups [
42]. The tendency that younger age was a much stronger risk factor for
vivax infection in the time period before the net distribution campaign should be interpreted with caution given the large and overlapping confidence intervals. The might be a topic of interest for future studies.
The study did not find evidence that gender, household wealth or education of head of household were associated with
P. vivax infection, which contrasts conclusions from other studies [
28,
31,
37,
40,
41]. Part of this discrepancy might be because the present study lacked power to detect these associations (versus Siri [
41]); categorized variables differently (versus Graves et al. and Khaireh et al. [
28,
31]; or required objective/subjective fever in addition to a positive RDT and microscopy reading to meet
P. vivax case definition (versus most other studies which did not require clinical signs or symptoms [
28,
31,
37,
41]. Lastly, given that the transmission of
P. vivax depends on many setting specific characteristics (e.g. altitude, rainfall, temperature), it might well be that a given variable is an important risk factor in one setting, though less important in another setting with different transmission patterns.
Study limitations
Loss to follow-up is a potentially important limitation to consider. However, because there was inconsistent evidence on how loss to follow-up was associated with proximity to breeding site, and because the design allowed for new participants midway through the study, it is difficult to evaluate how loss to follow-up might have affected results. There was consistent evidence that those lost to follow-up were younger and had a lower proportion with ITN use over 50% which would lead to an underestimation of the overall rate of infection in the study (the age difference was small in absolute terms, thus the likely biological effect of this should be limited). Since those lost to follow-up in addition had a higher proportion of males, care should be taken when generalizing to populations based on similar baseline characteristics.
The diagnosis of
P. vivax with microscopy is not straightforward in settings where
P. falciparum and
P. vivax coexist [
9,
17,
19], which could result in under diagnosis and/or misclassification of
P. vivax malaria (non-differential). In terms of ITN use, it is possible that reported use the night before the weekly visits was too imprecise to adequately control for ITN as a confounder, resulting in residual confounding. However, a strength of the study was the thoroughness of the data collection on ITN use compared to most other studies on
P. vivax malaria in the region. Residual confounding might also result from dichotomizing the ITN variable.
The decision that participants experiencing
P. vivax infection would exit the study at the point of their first infectious episode might also be considered a limitation. That is, by excluding the contributed follow-up time after infection for these participants, the final model used less information than was available in the data-set. Nonetheless, because ITN use patterns likely changed for participants as a result of being infected, and because repeat infections might be relapse episodes of the dormant hypnozoite stage of the
P. vivax parasite rather than truly new infections [
45‐
48], excluding participants upon their first episode was deemed the best approach.
Conclusion and recommendations
The present study found strong evidence for a positive association between P. vivax infection rates and living close to a vector breeding site. Contrary to earlier research, there was no evidence that this association varied across age groups. The transmission dynamics of P. vivax malaria depend on many factors, some of which are highly setting specific—e.g. climate and altitude. Therefore, care should be taken when trying to apply the findings from the present study to populations in different settings.
The findings might influence how net-distribution and indoor residual spraying campaigns are planned and implemented, help guide strategies on water resource development by highlighting potential negative health effects of man-made dams near human habitats, and add to current educational information given to people living close to vector breeding sites.
Authors’ contributions
AN: analysis and interpretation of data; drafting and revising manuscript; final approval. JC: analysis and interpretation of data; revising manuscript; final approval. EL and BL: conception and design of study; acquisition of data; revising manuscript; final approval. All authors read and approved the final manuscript.