Late incidence of SARS-CoV-2 infection in a highly-endemic remote rural village. A prospective population-based cohort study

ABSTRACT

Data on SARS-CoV-2 transmission in rural communities is scarce or non-existent. A previous cross-sectional study in middle-aged and older adults enrolled in the Atahualpa Project Cohort demonstrated that 45% of participants had SARS-CoV-2 antibodies, 77% of whom were symptomatic. Here, we assessed the incidence of SARS-CoV-2 infection in the above-mentioned rural population. One month after baseline testing, 362 of 370 initially seronegative individuals were re-tested to assess incidence of seroconversion and associated risk factors. Twenty-eight of them (7.7%) became seropositive. The overall incidence rate ratio was 7.4 per 100 person months of potential virus exposure (95% C.I.: 4.7–10.2). Six seroconverted individuals (21.4%) developed SARS-CoV-2-related symptomatology. The only covariate significantly associated with seroconversion was the use of an open latrine. Predictive margins showed that these individuals were 2.5 times more likely to be infected (95% C.I.: 1.03–6.1) than those using a flushing toilet. Therefore, along one month, approximately 8% of seronegative individuals became infected, even after almost half of the population was already seropositive. Nevertheless, a smaller proportion of incident cases were symptomatic (21% versus 77% of the earlier cases), and no deaths were recorded. Whether this decreased clinical expression resulted from a lower viral load in new infections cannot be determined. Increased seroconversion in individuals using latrines is consistent with a contributory role of fecal-oral transmission, although we cannot rule out the possibility that latrines are acting as a proxy for poverty or other unknown interacting variables.

Introduction

The Coronavirus Disease 2019 (COVID-19), caused by the novel Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) pandemic, has affected more than 30million people worldwide [1]. The disease has now spread to Latin American countries, claiming the lives of hundreds of thousands of people [2]. While individuals living in rural communities of these countries are supposed to be especially vulnerable to this pandemic because of factors inherent to under-development [3, 4, 8], data on SARS-CoV-2 transmission in these communities are scarce or non-existent.

Atahualpa – a rural Ecuadorian village – was severely struck by this pandemic, as noticed by a sudden increase in adult mortality associated with respiratory diseases during April and May, 2020 [9], coinciding with the introduction of SARS-CoV-2 in Ecuador [10]. We then conducted a study of SARS-CoV-2 prevalence in middle-aged and older adults enrolled in the Atahualpa Project Cohort, and demonstrated that almost half of adult individuals had SARS-CoV-2 antibodies [11]. In addition, a case-control study conducted in the village showed a strong household transmission of the disease between adults and younger individuals [12]. Taking the unique opportunity of this well-established cohort, we aimed to assess the incidence of SARS-CoV-2 infection in this rural population.

Material and methods

Study population

The study was conducted in Atahualpa, a remote rural village in coastal Ecuador, where previous studies on SARS-CoV-2 infection have been conducted [9,11–13]. Characteristics of Atahualpa residents have been described elsewhere [14]. The weather is hot and dry, with 12 daily hours of sunlight year-round. The village has electricity and almost all houses have piped water. However, about 20% of houses do not have a flushing toilet and still use open latrines [13]. The population is homogeneous regarding race/ethnicity, living conditions, and dietary habits; most men belong to the blue collar class (artisan carpenters) and most women are homemakers. In Atahualpa, there is only one health center of the Ministry of Health, and the nearest hospital is about 10 miles away in a small city (Ancón).

Background data

By May 2020, we conducted a cross-sectional study of SARS-CoV-2 infection in 673/691 (97.4% coverage) of middle-aged and older adults actively enrolled in the Atahualpa Project Cohort. In that study, 303 (45%) participants were positive for SARS-CoV-2 antibodies, 77% of whom were symptomatic [11]. It was also demonstrated a significant association between seropositivity and the use of latrines, suggesting a contributing role for fecal-oral transmission of the disease.

Study design

One month after the above-mentioned seroprevalence study, seronegative individuals were re-tested, using the same test for detection of SARS-CoV-2 antibodies (see below), to assess the incidence rate ratio (IRR) of seroconversion as well as associated risk factors. Poisson regression models were used to estimate factors associated with seroconversion. The study and informed consent were approved by the IRB of Universidad Espiritu Santo (FWA: 00028878), Guayaquil, Ecuador.

Serological tests

Detection of SARS-CoV-2 IgM and IgG antibodies was performed using the BIOHIT SARS-CoV-2 antibody test kit, colloidal gold method (BIOHIT Health Care Ltd., Cheshire, UK). In this immunochromatogaphic test, immunoglobulins from the sample react with SARS-CoV-2 recombinant antigens linked to colloidal gold particles and the resulting antigen-antibody complexes migrate along the membrane and react with mouse anti-human IgG or mouse anti-human IgM anti-antibodies. The manufacturer reports 97.5% sensitivity with 99.5% specificity for IgM, and 97.5% sensitivity with 100% specificity for IgG detection of this kit. IgM and IgG antibodies should become detectable during the 2nd week of the onset of symptoms, with IgM levels decreasing during the 4th week and IgG persisting beyond the 6th week or longer [15]. Digital photos of the tests were taken at the field, and independently read by two of the authors (OHD and BYR) blinded to any individual information. Discrepancies were resolved by consensus with the aid of a third reader (HHG).

Clinical examinations

Field personnel (including a medical doctor) visited houses of eligible candidates. Individuals were interviewed and examined regarding clinical manifestations related to COVID-19. World Health Organization (WHO) operational definitions of suspected cases were used, as follows: (1) acute respiratory illness (fever and at least one sign/symptom of respiratory disease, e.g. cough, shortness of breath) and a history of travel to or residence in a location reporting community transmission of COVID-19 during the 14 days prior to symptom onset; (2) any acute respiratory illness and having been in contact with a confirmed or probable COVID-19 case in the last 14 days prior to symptom onset; and (3) severe acute respiratory illness (fever and at least one sign/symptom of respiratory disease and requiring hospitalization) in the absence of an alternative diagnosis that fully explains the clinical presentation [16].

Covariates of interest

Covariates were selected if they have been suggested to play a role in disease acquisition or spread, or in the development of clinical manifestations [5, 6, 7, 17]. These included age, gender, level of education (primary school education or higher), alcohol intake (dichotomized in ≤50 or >50 g/day), number of persons living in the house, bedrooms per house, having an open latrine (as opposed to a flushing toilet system), home confinement during the previous two months, and cardiovascular risk factors (assessed yearly, last round in June 2019) by means of the American Heart Association (AHA) criteria [18]. According to the AHA, cardiovascular risk factors are defined when the subject is a current smoker, has a body mass index if ≥30 kg/m², has no moderate and vigorous activity, a non-healthy diet, blood pressure ≥140/90 mmHg, fasting glucose ≥126 mg/dL, or total cholesterol levels ≥240 mg/dL.

Statistical analysis

Data analyses were carried out by using STATA version 16 (College Station, TX, USA). In univariate analyses, continuous variables were compared by linear models and categorical variables by the x² or Fisher exact test as appropriate. Poisson regression models, adjusted for the above-mentioned covariates, were used to estimate the IRR of seroconversion, as well as factors associated with such incidence. Predictive margins (delta method) were used to estimate the incidence rate of seroconversion of variables reaching independent significance in Poisson regression models.

Results

From 370 individuals who were seronegative in the cross-sectional prevalence study, 362 took part in the present longitudinal study (the remaining eight declined consent). The mean age of the 362 participants was 59.9 ± 13 years, 205 (57%) were women, 71 (20%) disclosed heavy alcohol intake, and 190 (52%) had primary school education only. The mean number of individuals and bedrooms per house were 5.6 ± 3.2 and 2.6 ± 1.1, respectively. Sixty-three individuals (17%) used an open latrine (as opposed to a flushing toilet), and 142 (39%) had been confined to home during the previous two months. Twenty-one individuals (6%) were current smokers, 97 (27%) had a body mass index ≥30 kg/m², 30 (8%) had poor physical activity, 21 (6%) had a poor diet, 104 (39%) had blood pressure ≥140/90 mmHg, 74 (20%) had fasting glucose ≥126 mg/dL, and 33 (9%) had total cholesterol levels ≥240 mg/dL.

With the exception of home confinement, other preventive measures were difficult to assess because of illiteracy of the study population and cross-cultural factors. There was general awareness of the disease as people learned from this pandemic through television or newspapers. While more than 90% of the population used face masks, the same mask was frequently used for more than two weeks (because of costs) and masks were not properly used (often below the nostrils). Frequent hand washing with soap or alcohol was not a common practice, nor the use of surface disinfectants. Social distancing was not taken into account, however, the Ecuadorian government prohibited social meetings of more than 10 persons, closed bars and restaurants in the village, and imposed a curfew (from 2pm to 5am) since the first week of April.

Twenty-eight out of 362 participants (7.7%) became positive for SARS-CoV-2 antibodies, but only six of them (21.4%) had SARS-CoV-2-related symptomatology, including two who were hospitalized. None of the seroconverted individuals – symptomatic or asymptomatic – had died from the start of this survey (25 June 2020). Incident cases had a tendency to cluster around previously identified seropositive subjects, but several seroconverted individuals lived in blocks where other houses had a few or non-seropositive cases (Figure 1). In univariate analysis, seroconverted individuals used more frequently an open latrine than those who remained seronegative; however, the rest of covariates investigated did not reach significance (Table 1).

Figure 1.

Satellite view (Google Earth, Google Inc., Mountain View, CA) of Atahualpa. Red dots correspond to prevalent cases and yellow dots to incident cases. Incident cases have a tendency to cluster around previously identified seropositive subjects

Table 1.

Characteristics of 362 Atahualpa residents across categories of serological status (univariate analysis)

	Total series (n = 362)	Seropositive (n = 28)	Seronegative (n = 334)	p value
Age, years (mean ± SD)	59.9 ± 13	60.9 ± 13.1	59.8 ± 12.9	0.665
Female gender, n (%)	205 (57)	16 (57)	189 (57)	0.954
Heavy alcohol intake, n (%)	71 (20)	5 (18)	66 (20)	0.807
Primary school education, n (%)	190 (52)	18 (64)	172 (51)	0.193
Number of persons per house, mean ± SD	5.6 ± 3.2	5.8 ± 2.1	5.5 ± 3.3	0.637
Number of bedrooms per house, mean ± SD	2.6 ± 1.1	2.5 ± 1	2.6 ± 1.1	0.642
Use of open latrines, n (%)	63 (17)	9 (32)	54 (16)	0.032*
Confinement to home, n (%)	142 (39)	10 (36)	132 (40)	0.692
Current smoker, n (%)	21 (6)	2 (7)	19 (6)	0.672
Body mass index ≥30 kg/m2, n (%)	97 (27)	4 (14)	93 (28)	0.181
Poor physical activity, n (%)	30 (8)	2 (7)	28 (8)	0.819
Poor diet, n (%)	21 (6)	1 (4)	20 (6)	0.599
Blood pressure ≥140/90 mmHg, n (%)	104 (39)	9 (32)	95 (28)	0.678
Fasting glucose ≥126 mg/dL, n (%)	74 (20)	6 (21)	68 (20)	0.892
Total cholesterol ≥240 mg/dL, n (%)	33 (9)	5 (18)	28 (8)	0.094

*Statistically significant result.

The overall IRR was 7.4 per 100 person months of potential virus exposure (95% C.I.: 4.7–10.2). Poisson regression models adjusted for covariates of interest were used to estimate factors associated with seroconversion (Table 2). Again, the only variable remaining significantly associated with seroconversion was the use of an open latrine (as compared with a flushing toilet). The IRR for individuals using a flushing toilet system was 5.9 per 100 persons months (95% C.I.: 3.2–8.6), which increased to 15.3 per 100 persons month (95% C.I.: 4–26.7) for those using an open latrine. Predictive margins showed that individuals using a latrine were 2.5 times (95% C.I.: 1.03–6.1) more likely to be infected by SARS-CoV-2 than those with a flushing toilet.

Table 2.

Poisson regression model showing incident rate ratios (IRR) for seropositivity, after adjustment for all covariables

SARS-CoV-2 seropositivity	I.R.R.	95% C.I.	p value
Age	1.00	0.97–1.04	0.759
Female gender	1.04	0.39–2.73	0.943
Heavy alcohol intake	0.64	0.18–2.24	0.485
Primary school education	1.27	0.52–3.15	0.601
Number of persons per house	1.08	0.94–1.23	0.276
Number of bedrooms per house	0.98	0.59–1.32	0.532
Use of open latrines	2.58	1.06–6.31	0.038*
Confinement to home	0.68	0.27–1.71	0.417
Current smoker	2.03	0.40–10.3	0.393
Body mass index ≥30 kg/m2	0.50	0.17–1.50	0.217
Poor physical activity	1.14	0.26–4.97	0.866
Poor diet	0.50	0.06–3.96	0.511
Blood pressure ≥140/90 mmHg	1.28	0.53–3.10	0.585
Fasting glucose ≥126 mg/dL	1.13	0.43–2.96	0.798
Total cholesterol ≥240 mg/dL	2.25	0.82–6.15	0.115

*Statistically significant result.

Discussion

This prospective study, conducted in a cohort of seronegative community-dwelling middle-aged and older adults living in a rural Ecuadorian setting, demonstrates an overall IRR of 7.4 per 100 person months of potential virus exposure, despite a basal seroprevalence of 45%. These numbers confirm sustained transmission after a significant proportion of the population has been infected, and are in line with previous fears of mass spread of the disease in remote rural populations of Latin America [3–8],

Our findings are characteristic of remote populations where inhabitants are immunologically naïve to a rapidly spreading pathogen [19]. Despite the almost nil migration rate of Atahualpa residents, many men have to travel on a weekly basis to neighboring urban centers to trade the furniture they build as artisan carpenters. This was likely the path of entrance of the virus to the community. This problem has also been anticipated in Brazilian aborigines living in closed communities, since they will be forced to travel to highly infected cities to receive a compensatory monetary bonus from the government [8]. To the immunological naivete of these populations, it must be added the adverse circumstances in which they live regarding inadequacy of control measures, poor social determinants of health and inadequate access to medical care [13]. Rural communities of Latin America do not seem to be prepared for this situation, and will likely be one of the hot spots of this pandemic.

The burden and consequences of SARS-CoV-2 infection in remote rural settings are largely unknown. Indeed, we conducted a comprehensive literature search for publications on community-based epidemiological studies of SARS-Cov2 in developing countries (using relevant keywords), and identified no high-quality published population-based study assessing the incidence of SARS-CoV-2 infection in community-based settings in developing countries. We only found four preprints on serological surveys in Iran, South Africa and Brazil, all taking probabilistic samples in small fractions of the studied populations [20, 21, 22, 23]. However, the design of those studies does not allow to ascertain the incidence of this infection. In urban centers of developed countries, however, some prospective studies have focused on the increased rate of seroprevalence over weeks. For example, in a Swiss study – which sampled enrolled individuals weekly – the estimated seroprevalence increased from 4.8% in the first week to 10.8% in the fourth week of follow-up [24]. These numbers are not totally comparable with the present study, since we departed from an estimated seroprevalence of 45%. Moreover, that study only focused on the detection of IgG antibodies and not on the presence of IgG, IgM, or both.

In the present study, a smaller proportion of incident cases were symptomatic when compared with earlier cases (21% versus 77%), and no deaths were recorded. Whether this decreased clinical expression results from a lower viral load in new infections or to the fact that the initially seronegative population was not completely immunologically naïve to the virus, in the setting of a large subset of the population already infected, cannot be determined from our study [25,26].

Increased seroconversion in individuals using a latrine confirms our baseline seroprevalence findings, consistent with a contributory role of open latrines as a source of infection (fecal contamination) [11]. In addition, we found that a shared latrine was the most likely cause of incident SARS-CoV-2 contagion among several members of two otherwise unrelated families [27]. These findings are in line with previous studies that demonstrated the presence of the virus in human feces [28,29] as well as in rivers contaminated by wastewater [30]. However, we cannot rule out the possibility that latrines are acting as a proxy for unknown interacting variables. For example, several houses with open latrines, while having pipe water, receive water from faucets located in the backyards, and do not have an incorporated pipe water sink in the latrine cabin (as opposed to regular in-house restrooms where flushing toilets are most often located). This may discourage people to wash their hands after defecation.

Our study is the first to demonstrate incidence of SARS-CoV-2 infection and the factors associated with incident seropositivity in a remote rural village. Other strengths of this study include the high coverage and unbiased inclusion of long-term participants in the Atahualpa Project cohort, in whom several risk factors and conditions have been well characterized [14]. Previously retrieved information from these individuals will provide grounds for the conduction of ambispective studies aimed to assess long-term consequences of SARS-CoV-2 infections (cognitive decline, changes in the diameter of large intracranial arteries among survivors, psychological distress, and so on). Nevertheless, the study population was limited to individuals aged ≥40 years. As such, we missed the infection status of younger villagers and how it could have influenced the overall infectivity. The test we used is reported to be highly reliable, but we cannot rule out misclassifications due to false positive or false negative results [31], or the remote possibility of cross-reactions with other viral agents, i.e. dengue [32]. In addition, individuals recently infected could be in the process of building their antibody response and would have tested negative [15]. While the level of infection in Atahualpa could be similar to that of several villages in the region, it may be different to other realities in different Latin American countries. Further studies are needed to better estimate the overall impact of SARS-Cov-2 in rural settings.

Funding Statement

Study supported by Universidad Espíritu Santo – Ecuador

Disclosure statement

The authors declare no conflicts of interest to disclose.