Introduction

Cholera is an infectious disease caused by the Vibrio cholerae bacterium through the ingestion of faeces-contaminated water or food. Transmission is either direct through exposure to an infectious individual or indirect through faecal contamination in the broader environment. Humans are the only host but V. cholerae can also survive in open water [1]. Typical symptoms are sudden onset diarrhoea, stomach cramps and vomiting, leading to quick dehydration and potentially death if left untreated. About 20% of infected people become symptomatic, of which the majority remain mild and only 20% continue to develop severe symptoms. Case fatality among symptomatic patients can reach 50% in the absence of treatment but can be brought down to below 1% through early rehydration and, to a lesser extent, antibiotic treatment [2,3].

There are an estimated 1.3–4.0 million cholera cases each year globally, both in endemic areas as well as through outbreaks after introduction in non-endemic areas [4]. Cholera has been endemic in the East of the Democratic Republic of the Congo (DRC) at least since the mid-1970s and accounts for 5%-15% of all cholera cases worldwide [4]. Specific enabling factors such as civil war, breakdown of public infrastructure, high population movement and high population density aggravate other common deficiencies of the public health sector in DRC that enhance cholera outbreaks such as lack of waste water management and access to safe drinking water [5].

As in many Sub-Sahara African (SSA) countries, most cholera cases in DRC occur in rural or semi-rural areas confined to so-called “hotspots”, i.e. small geographical areas with high concentrated cholera incidence [6,7]. These hotspots are located mostly in the eastern part of DRC, where several major lakes function as environmental reservoirs. However, an epidemic that originated from Kindu (Maniema province, Eastern DRC) in mid-2015 slowly started progressing westwards along the major rivers during the course of 2016 and 2017, eventually spreading into 24 out of all DRC’s 26 provinces, many of which had been cholera-free for over 10 years [8]. This epidemic caused 55,000 cases and 1,190 deaths along its way, before reaching the capital of Kinshasa in November 2017. Over the next 3 months, more than 1,000 cases were recorded in Kinshasa with a case fatality ratio of 4% [9].

A number of general “classic” risk factors for cholera that are related to direct contact with diseased cases or faeces-contaminated food and water are well established, such as using open sources for drinking water; lack of latrine use, hand washing and personal hygiene measures; caring for symptomatic patients; and poor food protection and processing [10,11]. Most of this existing evidence originates from rural, semi-rural or peri-urban populations, and in particular from South Asia, while recent studies from outbreaks in SSA mega-cities such as Kinshasa remain scarce. However, cholera transmission dynamics are known to be different in urban SSA settings [12,13], and further urbanisation in combination with climate change will make urban cholera outbreaks more likely in these settings in the future [14,15]. At the same time, the World Health Organisation (WHO) has embarked on a roadmap to eliminate cholera in 20 countries by 2030 [7]. In DRC, the elimination of cholera was officially declared the aim of national health policy in 2007 and several strategic documents have been produced since then [16–18]. However, the resurgence and prolonged persistence of cholera in previously non-endemic areas of DRC shows that strategic planning alone does not eliminate cholera without, among other factors, having a good contextual understanding of transmission patterns to guide targeted outbreak control measures.

Therefore, it is paramount to have a better understanding of context-specific risk factors for cholera in Kinshasa and similar settings in SSA. We seized the 2017–2018 cholera outbreak in Kinshasa as an opportunity to investigate context-specific risk factors in a congested urban SSA setting among a population that is not frequently exposed to cholera to inform response to future outbreaks in similar contexts.

Methods

Results

After data cleaning and verification, we were able to use 390 observations from 195 pairs of 1:1 matched cases and controls for analysis, and constructed 20 variables from the information collected from CRFs. While the sex ratio was balanced between cases and controls, 225 (57.7%) of participants were female. While socio-economic characteristics did not differ substantially between cases and controls, univariate analysis suggested cases were more likely to have attended a funeral recently (7.5% vs 4.8%), less likely to not have consumed food at the roadside (19.8% vs 25.3%), and less likely to have washed fruit before consumption (40.3% vs 48.0%). Notably, more than 80% of all participants reported consumption of street vended sachet water. Controls were more likely to not have consumed this type of drinking water (23.5% vs 16.7%), while cases reported higher “often” or “very often” consumption than controls (52.1% vs 44.4%) (Table 1). Age, not a risk factor for analysis due to the matched study design, was evenly distributed among study participants with a mean of 23.1 years (SD 17.0), with 20–30 years olds accounting for 91 (23.0%) of all participants (S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Univariate analysis.

https://doi.org/10.1371/journal.pntd.0009477.t001

After applying ABE, 9 of the 20 variables were retained in the final conditional logistic regression model, of which only sachet drinking water consumption showed a strong association with the risk of being a cholera case (p-value 0.019). Among those who consumed street vended sachet water, the risk of cholera increased with increasing frequency of consumption from rarely [OR 2.2, 95% CI 0.9–5.2] to often [OR 4.0, 95% CI 1.6–9.9] to very often [OR = 4.1, CI 1.0–16.7]. Having attended a funeral and having had recent contact with a diarrhoea patient had a p-value of 0.09 and 0.08, respectively (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Multivariate analysis.

https://doi.org/10.1371/journal.pntd.0009477.t002

While 52.0% of participants had complete information on all variables, only three variables had more than 5% missing data. We obtained the same results for both methods of sensitivity analyses (Complete Case Missing Value and Subgroup Adjustment) as in multiple imputation (S3 and S4 Tables). No issues with multicollinearity between selected variables were detected (S5 Table).

Discussion

This was the first risk factor analysis for cholera in Kinshasa and one of few existing prospective studies from an urban Sub-Saharan African setting. We identified the consumption of street-vended sachet water as an important risk factor during the 2017–2018 Kinshasa outbreak. The detected dose-response effect, with increasing frequency of sachet water consumption being positively associated with increasing risk of cholera, together with the very high proportion of sachet water consumption overall makes us confident that this observed effect is real.

Our main finding might appear counter-intuitive at first sight since industrial processing of disinfected water into vacuum-sealed plastic sachets should eliminate most contamination risks, while untreated and unprotected open drinking water source is a well-established classical risk factor for cholera in numerous studies [10,11]. However, there is sporadic evidence that backs up our findings: Though V. cholerae has once been isolated from water sachets, albeit in low quantities, during a cholera outbreak in Nigeria in 2011 [24], it seems that sachet water quality at source or during processing and packaging is not the main concern. Several studies showed that, while bacterial contamination is not uncommon, the quality of the water inside the sachet is comparably higher and better for health than other common drinking water sources [25–28]. Instead, external contamination of the plastic sachet may be the mode of substantial pathogen transmission, as a recent study from Malawi suggests [29]. Water sachets are consumed by biting off an edge and sucking the water out of the sachet, thereby making direct and continuous contact between mouth and the sachet unavoidable. This mode of transmission was also hypothesized from an outbreak investigation in Accra, Ghana during a cholera outbreak in 2014. Among a population that relied heavily on sachet water as a source of drinking water, the investigators reported based on univariate analysis that consumption of this type of water was associated with a six-fold increase in cholera risk [30]. However, since no adjustment for potential confounding was done, the size of the reported effect should be interpreted with caution.

Funeral attendance and contact with symptomatic diarrhoea patients, variables with p-values between 0.1 and 0.05 in our study, are epidemiologically plausible risk factors for cholera infection and in line with other research [31,32]. Contact with a person showing symptoms of diarrhoea might have also functioned as a proxy of other unexplored risk factors in our study like shared contaminated water sources. On the other hand, other behavioural factors like place of food and fruit purchase or hygiene procedures prior to consumption that have been identified elsewhere [33] were not strongly associated with the outcome in our final model.

This was a field study carried out during an active outbreak, which led to certain limitations. First, not all included cases were laboratory-confirmed, and we do not know which ones. As per routine practice during cholera response, laboratory confirmation is stopped after an outbreak is detected and declared. This was also the case in Kinshasa, where only 117 sample from an overall 1,065 cases were sent for laboratory testing in the 2017–2018 outbreak [9]. Given the clear clinical picture of full-blown cholera, the fact that CTC admission criteria followed national and international standards [19,34], and that the triage staff were experienced health care workers should keep the proportion of false-positive inclusions among cases low. Second, taking stool samples from controls to rule out cholera infection was not feasible in our study, which might have led to an inclusion of some false inclusions among controls. Field enumerators followed standardized and detailed clinical questions to ascertain the eligibility of controls, which ruled out the inclusion of symptomatic persons among controls but not of asymptomatic infections. If present, both types of misclassifications would have led to an underestimation rather than an overestimation of the associations we detected in our study. Third, differentiating definitively between (internal) water contamination and (external) sachet contamination as mode of cholera transmission in this outbreak was not possible since this would have required systematic sampling and testing of water sachets along the production, retailing and consumption process. Forth, we did not asses vaccination status among participants from an oral cholera vaccination campaign done in 2016. However, given that most of our study area was not covered by that campaign and the fact that protective effects from cholera vaccines are not believed to last beyond two years, we believe that this did not affect our results substantially. Fifth, we only recruited 390 participants instead of our target sample size of 402, and some variables had missing data (Table 2). However, since the discrepancy between the actual and target sample size and the amount of missing data was not big, we do not believe that this affected our results substantially.

Noteworthy strengths of this study include prospective and systematic data collection using detailed SOPs and CRFs; the community-based design for recruitment of controls; the involvement of trained field epidemiologists as data collectors; and the application of advanced statistical methods to triangulate the validity of our findings.

Although cholera is an old disease with well-known classical risk factors, each setting is unique and produces its own context-specific transmission patterns and risk factors. Sachet water consumption has experienced a sharp rise during the past decade in SSA and is particularly common in cities rather than in rural areas [9,25,35,36]. With more than 80% of our study participants reporting the consumption of this type of water, Kinshasa is no exception to this phenomenon. Future behavioural cholera risk factors studies should ideally include mixed-methods approaches to contextualize findings and reveal other previously not considered modes of cholera transmission.

WHO’s 2030 roadmap, in which DRC features as a key country, and in particular DRC’s own cholera elimination agenda are very ambitious [7,16,37]. Early outbreak detection and quick response is the cornerstone of both strategies. For this, understanding local transmission patterns remains key, in particular in urban areas with particular risk behaviours for which existing evidence remains scarce. This is acknowledged in DRC’s current multisectoral strategy plan for the elimination of cholera, which explicitly recognises the need for continuous operational research to improve outbreak response interventions [17]. The use of oral cholera vaccination has gained momentum in recent years through the creation of a global stockpile and has, despite remaining challenges, become an important tool in the response to endemic and epidemic cholera [38,39]. However, as with other technical innovation, this must not come at the cost of traditional outbreak investigation and response measures that are informed through locally acquired contextual insights. For this, the role of field epidemiology will remain crucial. Only by adapting our variable selection and data collection tool to the local setting of Kinshasa and informed by local knowledge were able to identify sachet water consumption as an important risk factor in this outbreak.

Many parts of DRC had been free of cholera for many years. The nation-wide resurgence of cholera that started in 2015 in Eastern DRC, which ultimately led to the urban outbreak in Kinshasa described here, shows how easily gains can be lost. Our study highlights the need for and added value of tailored outbreak investigations as part of a comprehensive and successful response strategy.

In conclusion, the consumption of water from street-vended sachets needs to be considered as a potential route of cholera transmission in urban settings where the use of this drinking water type is frequent. Health messaging, outbreak control measures and epidemiological investigations should include the consumption of water sachets as a potential risk factors during future cholera outbreaks in urban low-resource contexts.

Supporting information

Acknowledgments

We thank all staff who were involved in this research and we are grateful to the patients who participated in this study.