In South Asia, kala-azar classically occurs in agricultural villages where houses are frequently constructed with mud walls and earthen floors, and cattle and other livestock are kept close to human dwellings. Nevertheless, the disease was reported historically from cities such as Calcutta and Madras [21],[22], and poor peri-urban communities in VL-endemic districts may still be vulnerable. The epidemiology of VL is characterized by both large-scale and small-scale spatial clustering. Despite the extensive geographical distribution of Ph. argentipes over much of the Indian subcontinent [23] and from Iran to Indonesia [24], VL cases are highly clustered in a small area of northeastern India, southeastern Nepal, and west and central Bangladesh. More than 90% of Indian VL cases occur in the state of Bihar and more than 50% of reported cases in Bangladesh come from 5 subdistricts of Mymensingh district representing less than 2% of the national population [17],[25].

At a much finer spatial scale, living close to a previous case of kala-azar strongly increases risk of both kala-azar and subclinical infection (Table 1) [7], [26]–[28]. In a study of one Bangladeshi community, people living within 50 meters had a 3-fold increase in kala-azar risk and those with a previous case in the same household were 26 times more likely to develop kala-azar, compared to individuals living more than 50 meters away [27]. These findings are thought to reflect the role of kala-azar patients as the most important infection reservoir during the study period. However, new data showing a recent increase in PKDL patients in Bangladesh highlight the likelihood that the role of kala-azar cases as the major leishmanial infection reservoir may not be constant over time [29]. Active case finding was conducted among 22,699 respondents in 2007–2008 in Mymensingh, including the area studied in 2000–2004 [29]. A total of 813 participants (3.6%) reported having had kala-azar since 2002, and 79 (9.7%) of these had developed PKDL by the time of survey, with a median interval of 21 months from kala-azar to onset of PKDL. Eight additional PKDL patients had no antecedent kala-azar history. The annual kala-azar incidence peaked at 85/10,000 in 2004, falling to 46/10,000 in 2007, but PKDL incidence rose steeply from 1/10,000 in 2004 to 21/10,000 in 2007. In Bangladesh, at least, it seems likely that PKDL patients will play an important role in sustaining transmission now and in the near future.

Only six explicit studies of risk factors for kala-azar and four for subclinical infection have been appeared in the international literature since the South Asian VL resurgence began in 1970 (Table 1) [7], [10], [26]–[28], [30]–[34]. The results of these studies should be viewed with caution, because most were cross-sectional studies, which by their nature cannot evaluate the effect of risk factors that could change over time. In the case of case-control studies, use of an appropriate control group is also essential to yielding valid assessments of risk. Nevertheless, a number of fairly consistent themes emerge, that can help inform thinking about interventions to control the disease. In the majority of studies, sleeping on the ground or outside was associated with increased risk of kala-azar, whereas sleeping on a cot was protective. Bed net use was protective against kala-azar in two studies [27],[30] but showed no significant effect in two others [26],[34]. Net condition, utilization prevalence, and effectiveness of insecticide impregnation, where applicable, may all affect the degree of protection associated with nets. Net use showed no protective effect against subclinical infection in 3 of 4 analyses [7],[32],[33]. In general, kala-azar risk was higher for children and young adults than for older adults who have a high prevalence of protective DTH responses [9]. However, in new foci where exposure is relatively recent, older adults may have an equally high risk of kala-azar [26]. Infants and young toddlers were relatively spared, perhaps due to being protected under a blanket or bed net during the hours of peak sand fly activity and consequently having less cumulative exposure [27].

Important environmental risk factors include living in a house with mud-plastered walls, especially if the walls are cracked; damp earthen floors; and closeness to small bodies of water and vegetation [30]–[33]. These factors are hypothesized to facilitate sand fly survival and increase vector abundance by providing breeding sites, diurnal resting places and humidity. The typical housing characteristics, as well as low educational level, lack of land, and other socioeconomic indicators, highlight the increased risk associated with poverty [26],[30],[35],[36]. Poverty mediates increased VL risk via multiple pathways, including increased sand fly access into poorly constructed houses and increased human exposure due to lack of personal protective measures such as nets [36]. However, within communities where characteristics of the local microenvironment are disproportionately influential, relative poverty may not be identified as a significant risk factor [27],[37].

Household members share more than just common exposures to potentially infected sand flies; genetic and nutritional factors may also be operative in this setting. In Brazil, specific gene loci coding for aspects of the immune response have been found to be associated with increased risk of symptomatic zoonotic VL in humans and dogs (the reservoir) [38]–[41], and poor nutritional status has been shown to increase the risk of progression from infection to kala-azar [42],[43]. Although similar data are not yet available for South Asian VL, both genetic and nutritional factors are assumed to play a role [7],[39]. Community-based data from Bangladesh demonstrated that population-level micronutrient status was poor, and higher (though still meager) intake of beef or goat increased the likelihood that an infection would remain asymptomatic rather than progressing to kala-azar [7]. Population status with respect to zinc, retinol, and other micronutrients known to modulate immunological status may be important in determining the proportion of infected individuals who become ill [7]. One study reported that non-leishmanial illness in the previous year increased kala-azar risk [31]; this factor may also act through a detrimental effect on immune status.

Data for the effect of cattle and other large livestock on human risk are contradictory, possibly reflecting the diversity of study designs and differences in statistical rigor. Ownership or proximity of livestock was associated with significant protection in some studies [27],[30], whereas, in others, kala-azar risk appeared to increase for those living in close proximity to cattle [26]. In Nepal, ownership of a cow or buffalo was associated with a significant decrease in kala-azar risk, but because of the case-control design, the authors were unable to conclude that protection was due to the entomological effects of cattle, as opposed to large livestock acting as an indicator of higher socio-economic status [30]. In the Bangladesh community-based study, cattle ownership was not associated with statistically significant protection, but each additional cow per 1000 m² around an individual's house decreased the risk of kala-azar by approximately 20% [27]. In other words, people were protected from kala-azar by their neighbors' cattle as well as their own. In the same study, the likelihood of a protective DTH response increased with increasing cattle density [7]. Interestingly, Napier and Das Gupta noted in 1931 that householders that kept cattle close to their dwelling places had a lower incidence of kala-azar [44]. A similar apparent contradiction is seen in African VL studies: a village-based study in eastern Sudan showed increased risk associated with cattle ownership for the first 3 years of a 5-year study [45], whereas a case-control study in Uganda indicated that treating livestock with topical insecticide against ectoparasites increased kala-azar risk, suggesting displacement of sand fly feeding to humans [46].

The complexity of the effect of cattle on VL risk may spring from the multiple ways in which their presence can alter sand fly density, infection rates and human-sand fly exposure, and to the complex and not yet fully elucidated population dynamics and behavior of Ph. argentipes (Table 2). Like the New World VL vector Lutzomyia longipalpis, Ph. argentipes is a predominantly peridomestic sand fly. Ph. argentipes breeds in moist organic soils at the junction of the floor and walls of cattle sheds and earthen houses [47]–[51]. Recent entomological trials of long-lasting insecticide-treated nets (LLINs) in Bihar failed to reduce indoor female Ph. argentipes densities, leading the authors to conclude that breeding may occur predominantly outside houses [52]. Ph. argentipes shows opportunistic feeding habits similar to those of Lu. longipalpis, whereby host “preference” is likely to be determined by host accessibility, abundance, size and biomass [53]. Ph. argentipes blood meals comprise predominantly bovine and human blood, influenced by trap positioning and trapping efficiency [54]. Although Ph. argentipes can be collected on bovine or human bait, several-fold more flies are generally caught on cattle than humans over the same time period [55] which probably reflects the influence of host density or biomass on sand fly feeding behaviour. As expected, blood meals in Ph. argentipes captured in cattle sheds and human dwellings reflect the predominant host (Table 2). Where humans and cattle are housed under the same roof, 19% and 66% of bloodmeals are of human and bovine origin, respectively [56]. In cattle sheds, 81–92% are bovine and 1–22% are human blood, and in human dwellings 49–100% are human and 0–39% are cattle blood [56]–[59], reflective of the complex dynamics of vector-host contact rates and consequences for transmission.

In Bihar, peak Ph. argentipes abundance has been documented in March-June (summer) and in October (just after the rainy season) and peak biting activity is seen late in the evening [55]. Ph. argentipes seasonality is similar in Bangladesh (unpublished data, Chowdhury, Dotson and Bern). Reports of Ph. argentipes “swarming” on cattle [60],[61], non-random spacing [60] and changing male to female ratios with time of night [55],[61] strongly suggest that Ph. argentipes forms leks for mating and blood feeding. This behavior and its implications for control interventions have been more fully described for the New World peridomestic sand fly Lu. longipalpis [54],[62]. Attracted by host kairomones, Lu. longipalpis males colonize the backs of animal hosts in the early evening, releasing long-range aggregation pheromones which attract more male and female flies [54],[62],[63]. For Lu. longipalpis, female recruitment reaches carrying capacity before that for males, such that the equilibrium male to female ratio is high [62]. Female blood feeding success (which relates to survival and fecundity) diminishes with increasing female fly aggregation density, which may then increase emigration to alternative aggregation sites [63]. Disrupting male recruitment pheromone production by insecticide application may lead to increased attraction to unsprayed sites or unprotected hosts. For example, spraying of animal sheds in Brazil resulted in increased sand fly densities in untreated dining huts [62], indicating that incomplete coverage of all aggregation sites could lead to increased sand fly abundance in vulnerable human dwellings and increased leishmanial transmission to humans. In summary, the impact of cattle on human kala-azar risk represents the outcome of a complex combination of effects on sand fly abundance, access and attraction to blood meal sources, and sand fly infection rates. Interventions that alter sand fly access to cattle should be carefully evaluated with respect to their impact on human exposure.

The studies reviewed in this article demonstrate that patterns of VL occurrence in South Asia are determined by the interplay of factors affecting sand fly abundance, infection rates and feeding behavior; proximity of infectious persons, especially kala-azar patients; population- and individual-level susceptibility; and determinants of human exposure to infected sand flies (Figure 1). This web of interactions offers opportunities for and lessons regarding control strategies (Boxes 1 and 2), and highlights fundamental gaps in the evidence base needed to promote the success of the current VL elimination initiative (Box 3).

The spatial patterns, especially at the community level, underscore the role of kala-azar patients as the major infection reservoir during periods of high disease incidence. Minimizing their infectious period through “rapid diagnosis and highly effective treatment” is appropriately a high priority within the South Asian control strategy, and the availability of new drugs and combination regimens has greatly improved the prospects for safe, successful kala-azar therapy [64]–[67]. However, PKDL treatment continues to rely primarily on courses of 120 intramuscular injections of sodium stibogluconate [68]. PKDL is less likely than kala-azar to prompt patients to seek care, because of the lack of systemic illness, and is difficult to treat effectively because of the prolonged drug course. If, as seems likely from xenodiagnosis studies, PKDL patients provide an important infection reservoir, the rising incidence of PKDL documented in Bangladesh poses a grave threat to VL elimination [13],[14],[29].

The relative importance of kala-azar, PKDL and asymptomatic infections as reservoirs in the transmission cycle is currently unknown. In one study in India, Leishmania parasites were visualized in peripheral blood smears of persons with asymptomatic infection [69]. In a Brazilian study, 25% of sand flies fed on kala-azar patients but none of those fed on asymptomatically infected humans became infected [70]. In Spain, HIV-L. infantum-co-infected patients were shown to be highly infectious to sand flies [71]; an increase in co-infection rates in South Asia would increase the effective infection reservoir. Based on sparse data from selected Indian patients, the proportion (8/8 PKDL, 15/18 kala-azar patients) that infected laboratory-reared Ph. argentipes and the percentage of sand flies infected (15–53%; 15–42%) were roughly similar for PKDL and kala-azar [13], [14], [72]–[77] (reviewed in [4]), but to date there have been no unbiased population studies of xenodiagnosis in anthroponotic VL.

Xenodiagnosis studies of dog populations are more numerous, and demonstrate that asymptomatic and presymptomatic L. infantum-infected dogs infect sand flies, but generally with lower frequency than dogs with symptomatic disease [78],[79]. Nevertheless, in some studies, a majority of asymptomatically infected dogs were infectious to sand flies [80],[81]. A recent meta-analysis demonstrates that the average percentage of dogs infectious to sand flies increased significantly with increasing clinical severity, from 29% of asymptomatic dogs to 80% of polysymptomatic dogs [4]. These findings, together with human epidemiological data, suggest that at least some asymptomatically infected humans may have the potential to transmit leishmaniasis, but are likely to be much less infectious than patients with active kala-azar or PKDL. However, at least 125 million people live in VL-endemic areas in South Asia, and perhaps 10 to 30% of these people have asymptomatic leishmanial infection. Even limited infectiousness of a fraction of these people could represent a large reservoir of infection and jeopardize efforts to eliminate the disease.

The failure of targeted vector control activities in South Asia since the 1990s suggests that vector programs for VL elimination need to be proactive, rather than reactive, and cover broad geographic areas. Current vector control strategies rely on IRS targeted to villages or parts of villages, in response to reported incidence of kala-azar in national passive surveillance systems [64]. Indeed, data from Bangladesh suggest that the delay between case detection and reactive IRS was so lengthy that it had no impact on transmission [27],[29]. The success of insecticide-treated nets against cutaneous leishmaniasis and sand fly vectors is well documented [82]–[88]. We hypothesize that VL control in South Asia could be more effective if IRS with broad coverage over VL-endemic areas is used to rapidly decrease sand fly populations and parasite transmission, and is simultaneously combined with active case finding and treatment, and widespread distribution of insecticide-treated nets to prevent the return of transmission from remaining kala-azar and PKDL patients when sand fly populations recover. Insecticide-treated nets could help fill the gaps that often result from the logistical difficulties of a sustained IRS program. Surveillance for HIV-VL coinfection and mechanisms to ensure that HIV-coinfected patients have access not just to rapid diagnosis and treatment of VL, but also antiretroviral therapy, secondary prophylaxis and insecticide-treated nets, will be necessary to ensure good clinical response and avoid an increase in the effective leishmanial infection reservoir. Human dwellings in South Asia are vulnerable to sand fly entry, even if breeding and aggregation occur predominantly in animal sheds. The effects of insecticide treatment of cattle sheds need to be evaluated, paying particular attention to the possibility that insecticide could repel sand flies resulting in higher biting rates on nearby unprotected humans. A simple intervention based on application of lime to mud walls showed promising entomological effects in a small pilot study [89]; further study to evaluate effectiveness and safety could add another inexpensive vector control method. Mitigation of the damp earthen floors and cracked mud walls of rural houses, and installation of screens or curtains to decrease sand fly access, could also play a role in decreasing human exposure to potentially infected sand flies, similar to housing improvement programs for control of Chagas disease [90].

Finally, crucial gaps remain in our knowledge base to achieve VL control (Box 3). The regional elimination program relies on the transient near-elimination of kala-azar during the malaria eradication era as a surrogate for proof of principle, while ignoring the rapid, explosive resurgence that occurred afterwards. Moreover, recent data are inadequate as a basis to design a control strategy with a high likelihood of lasting impact on VL incidence. Although a number of studies of the impact of insecticide-treated nets have been or are being conducted [91],[92], the resulting data may not provide a sufficient basis for elimination program policy decisions, given the complexity of the topic. As seen in the current review, the epidemiological determinants of kala-azar and subclinical infection are not necessarily the same; the use of seroconversion as the major outcome in an intervention trial may lead to erroneous conclusions about risk for disease. Similarly, sand fly density by itself does not predict infection risk; for example, successful CL control has been achieved in the absence of any measureable impact on the vector population [85]. In addition, no data are currently available to assess the duration of effectiveness of long-lasting impregnated nets under the conditions of usage in South Asian villages. To ensure the success of the initiative, it is more urgent than ever to conduct adequately powered, integrated demonstration projects to provide direct proof of principle for VL elimination. Several demonstration projects in carefully chosen sites may be necessary to ensure that the ultimate strategies are appropriate to the ecological and epidemiological characteristics of each location.

After kala-azar incidence is below the program threshold of 1/10,000 population, regional program documents propose a maintenance phase of 2–3 years [15],[64]. At the end of this period, sand fly populations are likely to rebound rapidly, as they did in the 1970s [16]–[20]. Based on the experience of the 1970s, maintenance activities will be necessary for a prolonged period of time, but currently there are no data to plan this critical phase. It is important to determine the most effective combination of maintenance activities and their necessary duration before national programs must make decisions on maintenance strategies and certification requirements. Finally, the resurgence in the 1970s was widely believed to have been sparked by PKDL patients who remained untreated for years [13], but the standard PKDL treatment regimen still requires 120 injections of a toxic drug. More effective, short-course treatment regimens for PKDL are urgently needed.

The South Asian regional initiative was begun with the admirable goal of eliminating the crushing burden of a lethal disease in some of the poorest populations of the world. Unfortunately, the evidence base to ensure the success of the initiative is still lacking. Without a coordinated, scientifically rigorous effort to fill these gaps in the near future, the current unparalleled opportunity to make a sustained impact on the disease burden caused by VL in South Asia may be lost.