Mechanisms of cholera transmission via environment in India and Bangladesh: state of the science review

Environmental transmission of cholera

The seasonal nature of cholera has been long recognised since the late 19th century. Globally, cholera epidemics demonstrate strong seasonal patterns in most endemic regions, with exceptions mostly located close to the equator where seasonal variation is limited [53]. Within the Indian subcontinent, two distinct patterns emerge. The first is a single peak during the rainy season (May-August) with often zero reported cases outside of seasonal outbreaks [17, 30]. This pattern is present in the drier parts of India including Delhi [54], Hyderabad [55], and Chandigarh [56]. The second pattern appears to be unique to the area surrounding the Bay of Bengal, namely West Bengal and Bangladesh. This is characterized by an annual dual-peak cycle; the first occurring in the spring (pre-monsoon) and the second in the Autumn (post-monsoon) with a marked lull in cases during the main monsoon [18, 57–61]. These two patterns suggest that cholera transmission is associated more with the environmental conditions of some seasons compared with others. Further, a 2011 study comparing spatial and social clustering of cholera in Matlab, Bangladesh found that while cholera always clustered spatially, clusters rarely occurred within social networks implying transmission was via environment rather than through person-to-person contact [62].

The evidence supporting a temporary environmental reservoir of pathogenic V. cholerae (strains within O1 and O139 serogroups) is strong. It has been frequently isolated during cholera outbreaks in a variety of fresh and brackish water sources across the Indian subcontinent [31, 63], [64], [65], [66], [67]. However, the onset of epidemics via primary transmission from an environmental reservoir requires persistence of pathogenic V. cholerae during inter-epidemic periods, a somewhat disputed topic in the literature [39]. In coastal regions surrounding the Bay of Bengal, the evidence is strong; both strains are considered autochthonous to the Gangetic Delta [66, 68, 69] and have been successfully isolated in coastal regions of Bangladesh between seasonal outbreaks [70] suggesting that V. cholerae are capable of surviving and even reproducing without human transmission. The permanent environmental reservoir theory explains the onset of seasonal epidemics by suggesting that V. cholerae concentrations increase in response to favourable conditions such as elevated water temperature [71], reduced salinity [23], and increased presence of phyto- and zooplankton which act as nutrient rich mini-reservoirs [66, 72].

However, other studies have shown that pathogenic V. cholerae require the presence of Na⁺ ions for survival [23, 24], with the O1 serogroup unable to survive beyond 24 h in salinity levels outside the Goldilocks range of 0.25–3.0% [73]. These findings are consistent with the proliferation of the bacteria in the brackish waters surrounding coastal regions, yet does not explain the persistent endemicity in many inland parts of the Indian subcontinent where the local aquatic environment contains almost zero salinity [31].

Miller et al. [74] proposed two potential explanations: (1) seasonal movements of people or availability of contaminated seafood; and (2) environmental reservoirs of pathogenic V. cholerae are in fact viable in freshwater environments e.g. through protection by aquatic biota such as zooplankton. More recent evidence supports the latter hypothesis, for example, a well-documented characteristic of cholera cases in Bangladesh is the simultaneous appearance of epidemics. This has been witnessed across both coastal and inland regions of Bangladesh [25, 75], as well as the simultaneous onset of mini-epidemic clusters within Matlab Demographic Surveillance Site in Bangladesh [19]. If the bacteria were spread only via movement of humans or food, a delay would be expected between outbreaks. The existence of simultaneous outbreaks therefore supports a strong role of primary transmission via the environment even in inland regions. However, Faruque et al. [76] suggest that while a delay between outbreaks may be present in the data, this is likely to be caused by a bias in reporting. They suggest that most outbreaks are caused by a single bacterial clone implying a single source, therefore arguing that the first few cases of a cholera epidemic go unnoticed in countries with limited health information infrastructure. If this is the case, separate outbreaks could indeed be spread from a single source, which would provide support for the first of Miller’s hypotheses.

With regard to the viability of inter-epidemic survival of V. cholerae in freshwater environments, Jubair et al. [77] propose the existence of ‘persister’ cells. The authors found that certain O1 V. cholerae bacteria were able to remain culturable for over 700 days in fresh lake water microcosms by switching to a ‘rugose’ phenotype. This switch causes the bacteria to produce Vibrio polysaccharide which confers protection against a variety of environmental stresses including osmotic and oxidative stress. However, the importance of these ‘persister’ phenotypes remains unclear as few published reports are available on the occurrence of rugose V. cholerae in the natural environment [78]. An alternative explanation is the capability of V. cholerae to enter a dormant low-metabolic state known as viable but non-culturable (VBNC) [79]. The bacteria enter this state in response to physiochemical stress (e.g. lower temperatures or nutrient availability), in which they do not respond to conventional culture methods of bacteria detection. Importantly, while it appears that V. cholerae are less infectious while in VBNC state [80], they are able to revert back to a culturable state upon animal passage [81] or the presence of more favourable growth conditions [82]. There is also evidence to support Miller and colleagues’ theory of protection from aquatic biota. Using fluorescent antibody techniques capable of detecting bacteria in VBNC state [83], found V. cholerae O1 survived for over 15 months in vitro in low salinity conditions (0.05%) within the mucilaginous sheath of the aquatic alga, Anabaena variabilis. The presence of VBNC V. cholerae O1 has been observed in situ outside of cholera outbreaks in freshwater samples from North India [67, 84], albeit in small concentrations. However, in a later study the authors sampled A. variabilis taken from pond water in Dhaka and were unable to isolate the bacteria during the inter-epidemic periods Jan/Feb and June/July [85]. This suggests that the persistence of pathogenic strains of V. cholerae during inter-epidemic periods in freshwater environments, while possible, is to be likely rare.

The precise mechanisms by which pathogenic V. cholerae is continually reintroduced into the environment, particularly in inland regions lacking brackish water, remains uncertain. It is nonetheless clear from the literature that secondary transmission in the region, which defines the growth, spread, and persistence of an outbreak occurs largely via the environment. This is made evident by the strong seasonal signature and frequent isolation of pathogenic bacteria in the aquatic environment during epidemics. As a result, in considering the factors that influence environmental transmission of cholera, this review will focus on secondary transmission.

Rainfall

The literature presented a complex relationship between cholera incidence and rainfall in India and Bangladesh. As mentioned in the introduction, the northern and inland endemic regions of India generally present a single peak in cases during the summer monsoon; several studies have demonstrated a positive correlation between the size of this peak and rainfall levels [30, 86]. Conversely, coastal regions surrounding the Bay of Bengal, namely Bangladesh and West Bengal, exhibit unique dual-peak behaviour with a lull in case numbers during the monsoon and peaks either side of the main rainy season. The second, and generally larger, of these peaks occurs post-monsoon and holds a similar positive association with rainfall [18, 57, 60]. Interestingly however, the first annual peak occurs in the pre-monsoon when the region is typically experiencing drought conditions. Further, during this period, cases are usually negatively correlated with rainfall, with drier years leading to stronger pre-monsoon peaks [18, 20]. In this section we provide the mechanisms presented in the literature to explain the both the positive and negative associations between rainfall and cholera incidence.

Ambient temperature

Laboratory microcosm studies have indicated a preference of V. cholerae for warmer waters, resulting in increased proliferation of the bacteria [23, 24]. This suggests an increase in ambient temperatures may result in a corresponding greater concentration of pathogenic V. cholerae and consequently greater probability that inhabitants will ingest an infectious dose. As a result, temperature has been invoked as a partial explanation for the Bangladesh spring peak as the rise in cholera cases coincides with elevated temperatures occurring between March and May [17].

The appearance of cholera cases in North India have been demonstrated to generally follow an increase in V. cholerae in the aquatic environment in the previous month, which is in turn preceded by a rise in temperature a further month before this [30]. Another study considering inland regions in Northern India/Pakistan found the odds of a cholera outbreak occurring were significantly higher when the temperature was above the climatological average over the previous 2 months [87].

Other studies have found less significant relationships between cholera incidence and ambient temperature. An autoregression study in Vellore [86] found no association, however lag times were not explicitly considered. A study in Bangladesh found a significant positive association between risk of cholera and heatwave after a 2-day lag, but only on wet days; on dry days there was no association [5]. They also found that the heatwave was effectively mitigated in households with a large percentage of medium-dense tree cover canopy around, potentially due to the cooling effects of shade and evapotranspiration. This suggests that rural areas may be less susceptible to heatwaves than urban areas with limited vegetation. Islam et al. [94] conducted a classification analysis on the relationship between climate variables and cholera cases in Matlab, Bangladesh. They found that increases in both temperature and sunshine hours positivity affect the variability of monthly cholera occurrence (i.e. high sunshine hours can compensate for low temperature). Summer in Bangladesh is associated both with higher temperature and fewer sunshine hours compared other seasons, and could therefore partially explain the summer lull in Bangladesh cholera despite the higher temperatures.

Ocean sea surface temperature (SST)

Ocean temperatures are unlikely to directly affect proliferation of V. cholerae as the bacteria do not live further from land than estuaries due to adverse conditions. However, due to coupled atmosphere-ocean processes such as the El Niño Southern Oscillation (ENSO) and the more localised Indian Ocean Dipole (IOD), ocean temperatures may influence cholera on land indirectly via changes in monsoon rainfall [95] and ambient temperatures [32] in India and Bangladesh.

A strong association between these climate processes and cholera have been extensively demonstrated in the literature in the context of the Bengal [28, 32], [33], [34], [35], [36], and was found to account for over 70% cholera variance in Dhaka during the period 1980–2001 [34]. However, studies describing the influence in inland India are more sparse. As with most influential factors, the strength of the association is dependent on context. Perez-Saez et al. [96] found spatial heterogeneity regarding its influence with a stronger ENSO sensitivity in the core of the city compared with the periphery. This matched similar results from Reiner et al. [97] who suggest the differences are likely to be caused by differences in population density and socio-economic conditions with the core of the city representing the highest levels of population density and poverty.

Coastal sea surface temperature (SST)

A related, though distinct, relationship appears to exists between coastal SST and outbreaks in Bangladesh. In 2000 Lobitz et al. [71] first found a statistically significant peak in cholera cases around a month following a peak in coastal SST for two out of the three years considered. The phenomenon has since been extensively studied, particularly among the coastal regions of the Bay of Bengal (BoB) however, findings have been inconclusive, as shown in Table 2. Some studies have concluded strong correlation between the two variables in line with the original findings from Lobitz et al. For example, a time-series regression by Hashizume et al. [35] found that each 0.1% increase in monthly SST in the preceding 0–3 months was associated with a 9.0% increase in cholera hospitalisations in Matlab and a 4.8% increase in Dhaka after accounting for sea surface height. Similarly, in Matlab, Ali et al. [3] found a 25% increase in cholera incidence associated with a one degree Celsius increase in SST in the current month, and 18% in the preceding month. When the spring and autumn peaks of the BoB were considered separately, Akanda et al. [57] found a strong positive correlation between summer SST and autumn cholera cases, and a strong negative correlation between winter SST and the spring cholera peak in Dhaka. Similarly Bouma and Pascual found a significant positive correlation with spring cholera in a Bengal dataset from 1891 to 1940, though found no association with winter/autumn cholera. Finally, others have found no significant correlation at all. Emch et al. [6] found no association when using an ordered probit model despite analysing a very similar dataset to Ali et al. a result repeated two years later by the same lead author [42]. De Magny et al. [68] developed predictive models to identify environmental signatures associated with cholera epidemics in both Kolkata and Matlab and found that SST was not a helpful predictor of cholera in either region. The authors suggested this was due to collinearity between variables.

The most obvious explanation for the positive correlation is the preference of V. cholerae for warmer water [3], as demonstrated in the previous section describing the positive association between ambient temperature and warmer waters. However, the inconsistency of this relationship suggests multiple mechanisms could be involved. The reviewed literature proposed two alternative theories as listed below.

Sea surface height

The reviewed literature contained two studies which directly considered the association between sea surface height (SSH) and cholera incidence. An early (2000) study from Lobitz et al. [71] found that cholera outbreaks in Bangladesh during the considered period were often preceded by an increase in SSH. However, only three years of data were considered and therefore statistical conclusions were unable to be drawn. A later statistical model which considered the influence of SSH on cholera hospitalizations in Matlab over a 24 year period found a significant negative correlation between SSH and cholera during the spring peak, but no association during the autumn peak [42].

Economic and social development

Unsurprisingly, the literature indicated a strong relationship between cholera incidence and socio-economic status. In a statistical analysis of cholera cases across India from 2010 to 2015, Ali et al. [109] found that % mobile phone ownership (used as a proxy for poverty) was a significant risk indicator for cholera, as were literacy rates. Further, Hamner et al. [110] found an increased likelihood of contracting cholera of 6.94 among families with monthly incomes under 3,000 rupees ($40) in Varanasi, India. The two studies which considered the relationship between socioeconomic status and hospitalisations in Bangladesh both focussed on the Matlab region and both found a significant association [42, 50].

Access and use of household water treatment systems are a potential mediator for this relationship. Point-of-use chlorine tablets have been shown effective at inactivating V. cholerae present in household water supplies in Dhaka [47]. Initial studies into the effectiveness of HWT systems on cholera incidence seem highly promising. An early intervention study in the urban slums of Calcutta, India found a 58% reduction in cholera prevalence when chlorinating water when compared to a control group [48].

An alternative mediating factor is population density which has been linked to cholera incidence in Matlab, Bangladesh [49]. This is likely to be due to increased contact between individuals, as well as the decreased sanitation conditions often found in dense urban areas. Within the rural area of Matlab, Study participants who shared latrines with other households had a 2.8 times great chance of being hospitalized with cholera [51].

Further socio-economic factors may include malnutrition which has been linked to prolonged shedding of V. cholerae [52]. Finally Education levels have been linked to both cholera incidence [50] and morbidity [49]. The precise mechanism by which education may affect cholera risk is not fully understood, but is considered to be related to personal hygiene [50].

Low levels of rain

The reviewed literature suggested three potential mechanisms for explaining the association between reduced rainfall and cholera incidence: reduced water volume, predation by vibriophages, and increased salinity. King et al. [111] have argued that this association is, in part, an illusion. They suggested that the monsoon lull is not due to environmental mechanisms but rather the inherently self-limiting nature of cholera outbreaks and high levels of asymptomatic cases that provide temporary immunity and deplete the susceptible population. During monsoon, the susceptible pool then replenishes, and is ready for another outbreak at the end of the season. However, Pascual et al. [112] question the validity of this theory by pointing out that seasonal outbreaks in Bangladesh tend to curtail prior to significant depletion of susceptibles. Further the depletion of susceptibles hypothesis cannot explain why drier years tend to result in stronger pre-monsoon peaks, and we therefore suggest that environmental mechanism(s) contribute to the associations between reduced rainfall and cholera incidence.

While the results of the review found evidence to suggest that each of the three proposed mechanisms was a plausible explanation of the drought-cholera association, none of the studies were able to demonstrate direct causal evidence making it difficult to differentiate between the relative importance of each. The comparative lack of reported drought-associated peaks within inland areas and the demonstrated association with conductivity in coastal regions perhaps points to an increased importance of the increased salinity hypothesis. However, this evidence merely suggests an association between coastal intrusion and cholera transmission could be caused by other coastal processes such as plankton abundance.

Finally, it is possible that the drought-cholera association can be explained by non-environmental processes. For example, stored household water is a significant source of V. cholerae contamination [48, 113], and therefore an increase in stored water may contribute to cholera outbreaks particularly during periods of drought. A study in Northwest Bangladesh found that households tend to store more water during the dry season due to uncertainties in supply [114]. Low availability of drinking water is also common during the dry season in coastal regions due to depletion and salinization of tube wells [115]. To our knowledge, no specific investigations have been conducted to test the hypothesis that increases in storage water act a mediating factor in the relationship between drought and cholera increase.

High levels of rainfall

The reviewed literature suggested two potential mechanisms for the association between increased rainfall and cholera incidence in inland India as well as the post-monsoon outbreak in the regions surrounding the Bay of Bengal: increased flooding, and contamination of water sources via wastewater runoff. Post-monsoon floods can be significant, particularly in Bangladesh where an average of 18% of land-surface is inundated and can even exceed 60% during high flood years [116].

In the literature we found conflicting results regarding the effectiveness of flood protection zones. An intervention in Dhaka suggested these zones conferred significant protection against cholera, however two studies considering the effects of a similar flood protection zone in Matlab suggested residents within the zone were at greater risk of cholera. This discrepancy could potentially highlight the difference in driving mechanisms between urban and rural areas. For example the flushing effect described by Carrel et al. could hold greater importance in rural areas and outweigh increased risks of cross-contamination. Whereas in urban areas, where population density is much higher, the risks of cross-contamination may be greater than that in urban areas. Further, risk of exposure to contaminated flood water may be greater due to lack spacing forcing households onto high risk areas. These studies may warn against the dangers of oversimplification when examining cholera transmission, as well as the importance of specific contexts.

Ambient temperature

The reviewed literature suggests that an association between ambient temperature and cholera incidence exists when considered at a lag time of around two months. Within the context of Northern India, Taneja et al. [30] note that elevated air temperature alone can only assist an already present reservoir of V. cholerae, and would not be sufficient to bring about an epidemic in regions with low or non-existence bacterial presence during inter-epidemic periods. In northern regions this would require the accompaniment of an appropriate transmission mechanism such as cross-contamination of surface water and sewage via increased rainfall. However, in more coastal regions where pathogenic V. cholerae are present year-round, elevated temperature may increase concentrations of pre-existing V. cholerae communities to sufficient level to cause infection.

Coastal sea surface temperature

The review found highly mixed results regarding the relationship between coastal SST and cholera outbreaks with different studies finding any of positive, negative or neutral relationships. Part of this complex relationship may be a result of the complexity of the relationship between phytoplankton and SST. Globally, it is generally the case that higher SSTs lead to reduced phytoplankton biomass due firstly to enhanced vertical stratification (resulting in reduced nutrient flux to the upper oceans where phytoplankton reside), and secondly an increase in the metabolic rate of primary consumers such as zooplankton [117, 118]. However, the negative effect of increased SST on phytoplankton productions appears to be strongest at higher latitudes. Gregg et al. [37] found that despite a global decline in ocean primary production in line with an increase in SST, production in the North Indian Ocean has increased by 13.6% between 1980 and 2002 – the most significant increase of all regions studied. This is supported by Chaturvedi [38] who found a year-round positive relationship between SST and chlorophyll in Northeast (coastal) Bay of Bengal. However, they found no relationship along the West coast of India. The implication of this is that while the association between SST and phytoplankton may be important to cholera prevalence in coastal Bangladesh, it is likely to be less relevant to outbreaks in India. Further, it is possible that the relationship between SST and phytoplankton may vary over time due to changes in global ocean processes. This may partially explain differences in associations with cholera when using colonial datasets compared with current, and may also suggest that extrapolating the results of this review when considering the effects of future climate on cholera incidence should be treated with caution.

A further complication in the relationship between SST and cholera incidence which may explain the inconsistent results is the presence of feedback loops within the proposed mechanisms influencing cholera incidence. Zooplankton rely on phytoplankton as a nutrient source, and thus their numbers are strongly linked in a predator-prey relationship [119]. Increased river runoff during the monsoon has been shown to dramatically increase phytoplankton biomass in the BoB [120]. Lewandowska et al. [118] have suggested that increased SST leads to the higher levels of chlorophyll in the BoB, but only when accompanied by major freshwater discharge. This fits with the findings from Coastal BoB, when seasonal river discharge is high, the correlation between SST and chlorophyll is positive. Coastal SST can affect rainfall over the subcontinent, and thereby influence cholera via rainfall mechanisms. A significant positive correlation exists between coastal SST in the Bay of Bengal and Bangladesh rainfall during the month of June [121]. This could potentially explain the strong positive correlation between summer SST and cholera prevalence in Dhaka during the autumn found by Akanda et al. [57] Our findings suggest the mechanisms mediating the association between coastal SST and cholera incidence are multiple and interlinked. Due to this complexity, the specific relationship between the two variables remains largely unclear, but is likely to be strongest in coastal regions during the spring peak.

Sea surface height

The results of the review produced limited evidence regarding the association between SSH and cholera incidence, although the available research suggests a neutral or potentially negative relationship. This is contrary to expectation as it could be anticipated that a positive relationship would be found due to coastal water intrusion or increased flooding. It has been argued that correlations found between SSH and cholera may be confounded by the third variable of increasing SST which mediates an increase in SSH through expansion of sea water [122]. Hashizume et al. [35] also found that an association between SSH and cholera was eliminated when adjusting for SST. Further, it should be noted that long-term indirect associations between SSH and cholera may become more apparent as gradual sea level rise caused by ice melt due to global warming leads to large scale population displacement and therefore increased population density [123], and saltwater contamination of crops and groundwater sources [124]. Further the GBM delta, which accounts for most of Bangladesh and the Indian state of West Bengal, is subsiding at a mean rate of 5.6 mm/year due to a variety of natural and anthropogenic causes [125]. This will further increase the apparent sea level rise and associated land loss in the region.

Economic and social development

While it is well established that poverty and cholera are strongly linked [126], the relative importance of each individual factor remains difficult to define. This is due to inherent and inevitable collinearity between poverty related variables. For example, despite it’s likely importance, sanitation conditions were not retained in a regression model to predict cholera incidence in Dhaka due to collinearity with population density [49]. Root et al. [50] found that some poverty related risk factors such as household assets and sanitation access were so highly correlated that separating the variables was a futile exercise. Nevertheless, given the faecal-oral nature of cholera transmission, certain factors such as access to adequate sanitation are widely accepted to be a fundamental cholera prevention strategy [127, 128].

Further we identified a significant lack of good quality studies assessing the effectiveness of specific HWT systems on cholera prevalence [129], particularly with other low cost treatments such as boiling which is highly effective at killing V. cholerae in laboratory environments [130]. The use of household water treatment (HWT) systems is limited within India and Bangladesh. According the 2012 India Human Development Survey, only 17.8% of those surveyed used any form of HWT such as boiling or filtration [131], while usage is even lower in Bangladesh, representing a mere 10.5% of the population in 2019 [7]. We therefore consider that there is significant potential for interventions in this area. Further research into the relative effectiveness of specific systems may allow increased efficiency of such interventions.

Limitations

This review has several limitations. Firstly the search was performed on only one platform and texts were selected for inclusion by only one researcher. Second, due to language limitations of the authors and an effort to limit the scope of the search, only texts published in English after 01/01/1970 were included in the search. This means that it is likely that some key texts may have been missed in the search process. Finally this review aimed to take a broad view at all evidence of factors which may influence environmental transmission of cholera across the relatively broad context of India and Bangladesh. For example, a systematic review focussed on the relationship between SST and cholera may help to disentangle the associated mechanisms and explain discrepancies between existing analyses.