Leptospira Infection in Rural Areas of Urabá Region, Colombia: A Prospective Study

ABSTRACT.

The objective of this study was to analyze the eco-epidemiological aspects of Leptospira seroprevalence and seroincidence and its associated factors in two municipalities of northwest Colombia. A prospective study was performed in rural areas of Urabá, Antioquia, Colombia. The study enrolled 597 people between November 2015 and January 2016, of which 274 were followed up 1 year later. Serologic testing was performed by a microscopic agglutination. The outcomes were seroprevalent and seroincident cases, and the primary exposure was an outdoor occupation. A binary and mixed-effect multinomial logistic regression model was used to estimate factors associated with seroprevalent or seroincident cases of Leptospira infection. The overall Leptospira seroprevalence was 27.81% (95% confidence interval: 23.62–32.49), and the overall cumulative seroincidence for Leptospira was 14.60% (95% confidence interval: 10.33–20.23). Multivariable analysis showed that factors associated with L. interrogans serogroups seropositivity were outdoor occupation, male gender, older age, the presence of dirt floor in the household, and the presence of piglets and opossums. It also showed that factors associated with other Leptospira species serogroups were the presence of pit latrines and of turkeys. In addition, the multivariable model of seroincident cases of L. interrogans serogroups evidenced outdoor occupations, the presence of rats, and corn cultivation as risk factors. Likewise, the multivariable model for seroincident cases of other Leptospira species showed that the presence of hunting canines and cassava cultivation were risk factors. We found specific factors associated with the transmission of Leptospira serogroups contribute to the understanding of the epidemiology of Leptospira infection in rural areas of Urabá, Colombia.

INTRODUCTION

Leptospirosis is a zoonotic bacterial disease that burdens resource-poor populations the most.¹ The disease is caused by pathogenic species of the genus Leptospira, which comprises 64 species.² The incidence rate of leptospirosis worldwide ranges from 0.10 to 975 cases per 100,000 population, and new cases are more frequent in tropical countries.^3,4 Environmental factors (rainfall, seasons, and erosions) and ecological factors (e.g., proximity of humans and domestic animals to wild mammals), and sociocultural characteristics (e.g., outdoor occupations and practices) increase the number of leptospirosis cases.^5–8

Leptospirosis has been a mandatory notifiable disease in Colombia since 2007,⁹ with a reported case fatality rate of between 1.40% and 2.24% during the period 2016–2018.^10–12 During the same period, 1,885 cases of the disease were confirmed; a significant proportion (18.30%) of these cases occurred in the Department of Antioquia, especially in the municipalities of Turbo and Apartadó in the Urabá region.^10–12 In this region, the seroprevalence of Leptospira was estimated to be 12.50% (95% confidence interval [CI]: 10.01–15.50) in 2007, with the highest seropositivity proportion found in Carepa (27.3%), Necoclí, and San Pedro de Urabá (25.00%), followed by Apartadó (14.80%), Turbo (11.80%), and Chigorodó (7.50%).¹³ However, higher seroprevalence was found later in Necoclí (35.60%) in 2009, indicating a potential increase in the transmission in the region.¹⁴ Among patients with febrile syndrome studied between 2007 and 2008 in Apartadó and Turbo, 31 cases (14.1%) were diagnosed with leptospirosis, and serogroups Tarassovi and Semaranga were identified in those patients.¹⁵ Moreover, studies conducted in the Urabá region have identified socioecological factors associated with L. interrogans exposure.¹⁴

Previous studies have shown that cases of leptospirosis are occurring in the Urabá region.¹⁶ The Urabá region, especially rural areas, has conditions favoring the circulation of Leptospira such as the presence of synanthropic rodents in households, the presence of wild animals in peridomiciliary areas, keeping domestic animals under unsanitary conditions, occupations with a potential risk of infection (agriculture), and inappropriate water supply and sewage disposal methods.^17–19 Furthermore, the failure to report cases of leptospirosis in Urabá has been related to poor recognition of areas and people at risk of presenting with the disease.^16,20 However, additional analytical studies aimed identifying factors associated with the seroprevalence and seroincidence of infection with different Leptospira serogroups have rarely been conducted in Colombia. The overall objective of this study was to analyze the eco-epidemiological aspects of Leptospira seroprevalence and the incidence and associated risk factors of infection in two municipalities of northwest Colombia. Our hypothesis was the socioecological factors (human, wild, and domestic animal proximity; age; occupation; land use; and household characteristics) are related to Leptospira seropositivity in humans.

MATERIALS AND METHODS

Study design.

A prospective study was conducted in two rural areas of Urabá, Colombia, in the localities of Alto de Mulatos, Turbo (8°08′12.5″N 76°33′01.7″W), and Las Changas, Necoclí (8°32′52.5″N 76°34′23.7”W) (Figure 1) located 356 and 418 km, respectively, from Medellín, the capital city of the Department of Antioquia. Between November 2015 and January 2016, individuals of both genders residing in the study area who agreed to sign the informed consent were included in the baseline study. People were excluded if they planned to move out from the study area within the next year and if they were suspected of being affiliated with illegal armed groups (this information was contributed by a research team that inhabited the study region). The same participants were followed up 12 months later, between November 2016 and January 2017.

Figure 1.

Map of the study area localized in rural areas of Urabá region of Antioquia, Colombia.

Sampling design.

Nine hamlets, five in Alto de Mulatos and four in Las Changas were included in this study. The characteristics of the population and study area have been described previously.²¹ Hamlet selection was made according to the ecological factors associated with infectious tropical diseases (the presence of wild mammals, mosquitos, and ticks, among others potential vectors and amplifying host animals), and sociodemographic characteristic such as hamlets with > 10 households, no illegal armed groups present in the area, and short distance to urban centers (< 2 hours travel time by horse) to facilitate study logistics. This study constitutes a secondary prospective analysis of the main study designed to estimate the seroprevalence against agents of the Rickettsia genus.²² However, the seroprevalence used for sample size estimation and the sampling design allowed other outcomes to be evaluated with confidence. A cluster random sampling was carried out with households as the sample units and their inhabitants as the analysis units. From the 461 households and 1,915 inhabitants identified in the census tract, a sample size of 208 households was estimated, using a 95% level of confidence, 5% error, and 41% expected prevalence.²² The sampling was proportional to the number of households in each hamlet, and a second serum sample was collected a year later from the same individuals included in the baseline study.

Leptospira seropositivity.

The research team collected blood samples from participants at baseline and at follow-up. The serum samples were tested by microscopic agglutination test (MAT). To select the antigens for serum samples testing, a panel of 32 Leptospira strains, representing different species and serogroups, was evaluated (Table 1) using a random selection of 20% of the total human serum samples collected at baseline. After the evaluation, antigens with no agglutination results were eliminated from the panel, and a final list of 16 antigens was established to be tested with all sera for this study (Table 1). The MAT assay was performed according to Felzemburgh et al.⁶ The evaluation and final screening assays were performed using dilutions of 1:50 and 1:100 for human samples, and a positive sample was determined when 50% or more leptospires for a specific strain were agglutinated. After the screening assay, positive samples were titrated to determine the higher titer for each strain.

Table 1.

Panel of Leptospira strains used in this study for the microagglutination test

Species	Serogroup	Serovar	Strain
L. interrogans	Djasiman	Djasiman	Djasiman
L. interrogans	Icterohaemorrhagiae	Icterohaemorrhagiae	RGA
L. interrogans	Icterohaemorrhagiae	Copenhageni	M 20
L. interrogans	Icterohaemorrhagiae	Copenhageni	L1 130
L. interrogans	Sejroe	Hardjo	Hardjoprajitno
L. interrogans	Sejroe	Wolffi	3705
L. interrogans	Bataviae	Bataviae	Van Tienen
L. interrogans	Pyrogenes	Pyrogenes	Salinem
L. interrogans	Pyrogenes	Manilae	L495
L. interrogans	Pomona	Pomona	Pomona
L. interrogans	Autumnalis	Autumnalis	Akiyami A
L. interrogans	Canicola	Canicola	H. Ultrecht IV
L. interrogans	Hebdomadis	Hebdomadis	Hebdomadis
L. interrogans	Australis	Bratislava	Jez Bratislava
L. noguchii	Louisiana	Louisiana	LSU 1945
L. noguchii	Panama	Panama	CZ 214 K
L. borgpetersenii	Tarassovi	Tarassovi	Perepelitsin
L. borgpetersenii	Ballum	Castellonis	Castellon 3
L. borgpetersenii	Ballum	Ballum	Mus 127
L. borgpetersenii	Mini	Mini	Sari
L. weilii	Celledoni	Celledoni	Celledoni
L. weilii	Javanica	Coxi	Cox
L. kirschneri	Cynopteri	Cynopteri	3522C
L. kirschneri	Grippotyphosa	Grippotyphosa	Duyster
L. santarosai	Shermani	Shermani	1342 K
L. santarosai	Tarassovi	ND	AIM
L. santarosai	Tarassovi	ND	JET
L. alexanderi	Manhao	Manhao 3	L 60T
L. alstoni	Ranarum	Pingchang	80-412T
L. kmetyi	Tarassovi	Malaysia	Bejo-Iso9
L. mayottensis	ND	ND	200901122
L. biflexa	Semaranga	Patoc	Patoc 1

Final selection of strains used for microagglutination test evaluation of all human serum samples are shown in bold.

Outcome definitions.

In the primary analysis, the outcome of seroprevalent cases of Leptospira infection was defined as a positive MAT titer ≥ 1:50 with 50% or more leptospires agglutination at baseline. The outcome of seroincident cases of Leptospira infection was measured by MAT seroconversion from negative to positive or by a minimum 4-fold increase in titers between baseline and follow-up samples. The serogroup that represented the highest antibody titer was defined as the presumptive infecting serogroup. However, when two or more serogroups had the same titer, the sample was considered mixed serogroup seropositivity. The seroprevalent and seroincident outcomes were categorized as seropositive to Leptospira interrogans serogroups seropositive to other Leptospira species serogroups and seronegative to Leptospira to perform a secondary analysis (multinomial analysis) (Table 1).

Data collection, study exposures, and covariates.

We adapted the malaria questionnaire used by Turner et al.²³ for our epidemiological survey. The questionnaire was adapted by the study team, which included anthropologists, local physicians, epidemiologists, health professionals, and members of the community. The research assistants, who are residents in the study area, contributed to adjustment of the questionnaire. A total of four assistants, two nurses and two previously trained individuals from the community, carried out the pilot study in a small group of residents of the urban center of Necoclí and Turbo to verify that study participants understood all the questions. We administrated two questionnaires to the study participants. The first was administrated to each individual and included 27 items related to sociodemographic and ecological variables. The second was administrated to the household head and included 18 items related to household characteristics such as wall, floor and roof materials; peridomicilliary characteristics (10 meters around the house); and presence of domestic, synanthropic, and wild animals, as well as practices related to risk factors for tropical infectious diseases.

The main exposure was working outdoors (farmers, agricultural workers, day laborers, and water collectors, etc.). For the baseline study, this exposure was analyzed as a previous outdoor occupation (within the past 5 years) and a recent outdoor occupation (within the previous year before the sample collection). For the prospective study, outdoor occupation was evaluated based on the previous year before the sample collection.^3,22–25

Additionally, individual variables such as age (years), gender, ethnicity, time of residence in the study area, educational level, and previous episodes of fever (any past episode and during the last year of follow-up) were evaluated as potential confounders of the association between the main exposure and the outcomes.

Secondary exposures were access to public services (the presence of indoor plumbing, waste disposal, the presence of sewage, the presence of latrines, and the presence of pit latrines), the presence of domestic animals in intra- or peridomiciliary areas (the presence of canines, felines, poultry, turkeys, pigs, horses, donkeys, and mules), and the presence of synanthropic or wild animals in peridomiciliary areas (the presence of rodents and opossums). Covariates such as the type of roof and floor, wall materials, the characteristics of peridomiciliary areas (the presence of bushes, trees, grasses, corn cultivation, cassava cultivation, and tomato cultivation), household location (urban center and rural area) and household proximity (very near, near, scattered, and very scattered) were analyzed. We used the following definitions for household proximity: very near when houses are attached to each other; near when households are located in front or by the side, but not attached; scattered when houses are dispersed but visible; very scattered when houses are dispersed and not visible. Finally, common practices among family members such as forest fragmentation or deforestation, and the use of any rodent elimination measures were considered as covariates. Forest fragmentation or deforestation was defined as practices related to cutting down trees in the forest around the study area and using those areas for culture.

Statistical analysis.

The seroprevalence and cumulative seroincidence were estimated considering the number of seropositive cases in the numerator and the number of people evaluated in both times (baseline and follow-up study) as a denominator. The confidence intervals of the seroprevalence and seroincidence were adjusted by random effects of the models (hamlets). In addition, the seroprevalent and seroincident cases were characterized using relative and absolute frequencies for qualitative variables and median and interquartile range for quantitative variables.

To estimate risk factors for seroprevalent and seroincident cases of Leptospira infection, a mixed-effects binary logistic regression model was used (seronegative versus seropositive samples). A multinomial analysis was conducted using a mixed-effects multinomial logistic regression model for a specific evaluation of factors associated with the outcome of seropositive against L. interrogans serogroups and seropositive against other Leptospira species serogroups. All models included three levels: individuals within households, households within hamlets, and hamlets (a random effect model using a variance component correlation matrix). The association between the main outcome (Leptospira seropositivity in humans) and variables at each level was evaluated (household and hamlet level). The linearity assumption was confirmed before the inclusion of quantitative variables in bivariate and multivariable models. Variables included in the multivariable models were those with P < 0.25 in bivariate analysis. The multivariable analyses were performed using the stepwise method based on the purposeful selection of variables, including both statistical and researcher’s criteria (biological plausibility).²⁶ To account for differential risks among different outcomes in the multinomial model, we built a specific model for each of the L. interrogans and other Leptospira species serogroup outcomes. The multilevel models of risk factors for Leptospira seropositivity in humans were weighted by the inverse probability of human selection in each hamlet. Confounding and effect modification were evaluated in multivariable models, and the model that best explained the outcome was selected according to Bayesian information criteria. Odds ratios (OR) were estimated in the cross-sectional (baseline) study, and relative risks (RR) were estimated in the prospective (follow-up) study according to Localio.²⁷ All analyses were performed in SAS 9.04.01 using PROC GLIMMIX.

Ethics statement.

The Institutional Review Board of the University of Antioquia approved all procedures carried out in the present study, and participants were enrolled according to written informed consent procedures previously approved.

RESULTS

Leptospira seroprevalence and associated factors.

Binary outcome analysis.

For the baseline study, 597 individuals (255 from Alto de Mulatos and 342 from Las Changas) that inhabited 246 households located in the rural areas (18% of oversampling) were enrolled. One hundred and three of them were in Alto de Mulatos and 143 in Las Changas. The sampling coverage of households and persons were 100% and 58.41% (597/1,022), respectively. The overall Leptospira seroprevalence was 27.81% (166/597) (95% CI: 23.62–32.49), and the socioecological characteristics showed that 42.77% (71/166) of seropositive individuals had outdoor occupations, 52.41% (87/166) identified themselves as of male gender, and the mean age in years was 31.43 (interquartile range [IQR]: 18.94–48.72). Among the seropositive individuals, 62.65% (104/166) reported a history of fever, and 79.52% (132/166) reported inhabiting houses with dirt floors. Regarding the presence of domestic animals in intra- or peridomiciliary areas, 55.42% (92/166) of seropositive individuals reported having seen opossums in the peridomiciliary, area and 26.51% (44/166) reported the presence of piglets in their household Supplemental Table 1).

Our analysis including the effect modification of occupation by gender identified the presence of a dirt floor in the household (OR = 1.77; 95% CI: 1.24–2.52), and the tenancy of piglets (OR = 1.52; 95% CI: 1.07–2.17) as risk factors to seropositivity against Leptospira. Also, age as a categorized variable (OR _{> 15–29 years versus ≤ 1–15 years} = 2.12; 95% CI: 1.35–3.32 and OR _{> 46 years versus ≤ 1–15 years} = 1.65; 95% CI: 1.02–2.69), the male gender (OR = 1.91; 95% CI: 1.33–2.73) and a history of fever (OR = 1.42; 95% CI: 1.05–1.91) were factors associated with seropositivity Supplemental Table 1). The stratified analysis by gender showed that the presence of piglets (OR = 1.68; 95% CI: 1.05–2.68) was the only risk factor in females. In males, factors associated to Leptospira seropositivity were older age (OR _{> 15–29 years versus ≤ 1–15 years} = 3.55; 95% CI: 1.73–7.27 and OR _{> 29–46 years versus ≤ 1–15 years} = 2.73; 95% CI: 1.24–6.01), outdoor occupation (OR = 2.03; 95% CI: 1.11–3.69), history of fever (OR = 1.78; 95% CI: 1.13–2.80), and dirt floor in households (OR = 1.96; 95% CI: 1.19–3.23).

Multinomial outcome analysis.

According to our multinomial analysis, the seroprevalence of L. interrogans serogroups was 18.25% (109/597) (95% CI: 14.34–23.11), and the seroprevalence of other Leptospira species serogroups was 9.56% (57/597) (95% CI: 7.11–12.70). Among the participants seropositive to L. interrogans, the most frequent serogroup was Bataviae, followed by mixed serogroups and Djasiman. Among the participants seropositive to other Leptospira species, the most frequent serogroup was Tarassovi, followed by Panama and Cynopteri (Figure 2).

Figure 2.

Frequency of Leptospira interrogans and other Leptospira serogroups in the seroprevalent cases.

Sociodemographic characteristics showed that 58.72% (64/109) of seropositive individuals against L. interrogans and 40.35% (23/57) seropositive for other Leptospira species were male. The median age of seropositive individuals was 33.23 years (IQR: 19.67–49.25) and 28.42 years (IQR: 16.76–44.66) among L. interrogans and other Leptospira species, respectively. Of note, a higher proportion of the seropositive to L. interrogans serogroups had an outdoor occupation (52.29%, 57/109), whereas only 24.56% (14/57) of seropositive persons to other Leptospira species serogroups had similar occupations Supplemental Table 2).

Most of the participants that were seropositive to L. interrogans (81.65%, 89/109) and other Leptospira species serogroups (75.44%, 43/57) inhabited households with dirt floors. Approximately half of the seropositive individuals in both groups had indoor plumbing (49.54% for L. interrogans and 54.39% for other Leptospira species). Additionally, the presence of pit latrines in the households was 45.78% (52/109) for people with L. interrogans seropositivity and 47.37% (27/57) for other Leptospira species seropositivity Supplemental Table 2).

Regarding the presence of domestic, synanthropic, and wild animals, 60.55% (66/109) of people with seropositivity against L. interrogans, and 45.61% (26/57) of people with seropositivity against other Leptospira species reported the presence of opossums in the peridomiciliary area. The presence of rats in intra- and peridomiciliary areas was 84.40% (92/109) and 85.96% (49/57) in people seropositive against L. interrogans and other Leptospira species serogroups, respectively. In addition, 1.83% (2/109) of people seropositive to L. interrogans and 7.02% (4/57) of people seropositive to other Leptospira species had hunting canines Supplemental Table 2).

The multivariable multinomial analysis including the effect modification term occupation by gender showed that occupations such as ranching, daily labor, farming, and agricultural (or similar) were not associated to the outcome of L. interrogans serogroups seropositivity (OR = 1.54; 95% CI: 0.89–2.67). However, male gender (OR = 2.02; 95% CI: 1.27–3.21), young age (OR _{> 15–29 years versus ≤ 1–15 years} = 2.16; 95% CI: 1.23–3.79), the presence of a dirt floor in the household (OR = 2.03; 95% CI: 1.31–3.16), and the presence of piglets (OR = 1.64; 95% CI: 1.09–2.47) and opossums (OR = 1.51; 95% CI: 1.06–2.15) in the peridomiciliary area were factors associated with L. interrogans serogroups seropositivity (Table 2). We then analyzed the effect modification, stratifying the model by gender and found that the outdoor occupation was a risk factor in males (OR = 4.76; 95% CI: 2.22–10.23), but not in females (OR = 0.74; 95% CI: 0.34–1.65). Age between 15 and 29 and > 46 years (OR _{> 15–29 years versus ≤ 1–15 years} = 2.41; 95% CI: 1.09–5.31 and OR _{> 49 years versus ≤ 1–15 years} = 2.48; 95% CI: 1.10–5.60), as well as the presence of piglets (OR = 2.36; 95% CI:1.34–4.14) were factors associated with leptospiral infection only in females. Households with dirty soil were a risk factor in both males and females (OR_females = 2.53; 95% CI: 1.18–5.41 and OR_males = 1.94; 95% CI:1.10–3.43) and presence of opossums was not associated with leptospiral infection in males or females (Table 3).

Table 2.

Multivariate analysis of seroprevalent cases of Leptospira among different serogroups (mixed-effects multinomial logistic regression model)

Variables	L. interrogans serogroup	Others Leptospira species serogroups
	ORadjust (95% CI)*	ORadjust (95% CI)
Individuals
Outdoor occupation†	1.54 (0.89–2.67)	0.88 (0.47–1.65)
Male gender	2.02 (1.27–3.21)	1.53 (0.90–2.60)
Age, years (1–15)	1.00	1.00
Age, years (> 15–29)	2.16 (1.23–3.79)	1.79 (0.95–3.36)
Age, years (> 29–46)	1.47 (0.81–2.69)	1.09 (0.54–2.21)
Age, years (> 46)	1.58 (0.86–2.90)	1.54 (0.77–3.06)
Household characteristics
Floor material
Presence of dirt floor†	2.03 (1.31–3.16)
Public services in household
Presence of pit latrine†		2.34 (1.27–4.32)
Presence of wild animals in peri domiciliary area
Presence of opossum†	1.51 (1.06–2.15)
Animal husbandry purpose
Presence of piglet†	1.64 (1.09–2.47)
Presence of turkeys (domestic animal)†		4.99 (2.03–12.30)

CI = confidence interval; OR = odds ratio.

^*Model including the effect modification occupation × gender.

^†The reference category is the absence of the characteristic evaluated.

Table 3.

Multivariate analysis of seroprevalent cases of L. interrogans serogroups stratified by gender

	Females N = 364	Males N = 233
Variables	ORadjust (95% CI)	ORadjust (95% CI)
Individuals
Outdoor occupation‡	0.74 (0.34–1.65)	4.76 (2.22–10.23)
Age, years (1–15)	1.00	1.00
Age, years (> 15–29)	2.41 (1.09–5.31)	1.92 (0.81–4.56)
Age, years (> 29–46)	1.62 (0.70–3.74)	1.38 (0.53–3–57)
Age, years (> 46)	2.48 (1.10 −5.60)	0.99 (0.38–2.61)
Household characteristics
Floor material
Presence of dirt floor‡	2.53 (1.18–5.41)	1.94 (1.10–3.43)
Presence of wild animals in peri domiciliary area
Presence of opossum	1.49 (0.90–2.49)*	1.54 (0.94–2.56)†
Animal husbandry purpose
Presence of piglet‡	2.36 (1.34–4.14)	1.13 (0.62–2.07)

^*missing data (N = 4).

^†missing data (N = 3).

^‡The reference category is the absence of the characteristic evaluated.

The multivariable model for the outcome of seropositivity to other Leptospira species serogroups adjusted by occupation, gender, and age showed that the presence of pit latrines (OR = 2.34; 95% CI: 1.27–4.32) and turkeys as domestic animals (OR = 4.99; 95% CI: 2.03–12.30) were risk factors (Table 2). We did not find any effect modification between occupation and gender in this outcome.

Leptospira seroincidence and associated factors.

Binary outcome analysis.

In the follow-up study, 274 participants were enrolled at 12 months; 120 were followed up in Alto de Mulatos and 154 in Las Changas. The characteristics of the people included in the study and people lost to the follow-up in both areas are shown in Supplemental Table 3. The overall cumulative seroincidence of Leptospira was 14.60% (40/274) (95% CI: 10.33–20.23), 35.00% (14/40) of the seroincident cases had an outdoor occupation, and 27.50% (11/40) were identified as male. The median age in years of seroincident cases was 40.30 (IQR: 20.46–55.34), and 80.00% (32/40) were inhabitants of a household with a dirt floor. Furthermore, among the seroincident cases, 5.00% (2/40), 2.99% (3/40), and 85.00% (34/49) of the seroincident cases disclosed the presence of corn cultivation around the household and hunting canines and rats in the intra or peridomiciliary areas, respectively Supplemental Table 4). According to risk factors associated with seroincident cases of Leptospira infection, an outdoor occupation in the past year (RR = 2.66; 95% CI: 1.46–4.68), a dirt floor in the household (RR = 1.77; 95% CI: 1.01–3.06), the presence of corn cultivation in the peridomiciliary area (RR = 8.33.74; 95% CI: 2.21–19.65), the presence of hunting canines (RR = 4.78; 95% CI: 1.83–10.74), and rat infestation (RR = 2.00; 95% CI: 1.03–3.76) in the intra-220 or peridomiciliary area were identified Supplemental Table 4).

Multinomial outcome analysis.

The multinomial analysis regarding serogroups showed that the seroincidence for L. interrogans serogroups and other Leptospira species serogroups was 8.39% (23/274) (95% CI: 5.25–13.15) and 6.20% (17/274) (95% CI: 3.38–10.93), respectively. Seroincident cases of L. interrogans and other Leptospira serogroups were male in 30.43% (7/23) and 23.53% (4/17) of cases, respectively. Additionally, 39.13% (9/23) of seroincident cases of L. interrogans serogroups and 29.41% (5/17) of seroincident cases of other Leptospira serogroups had had outdoor occupations in the past year. The median age for the seroincident cases was 47.46 years (IQR: 24.76–58.42) and 29.99 years (18.19–49.28) for L. interrogans serogroups and other Leptospira species serogroups, respectively. Only one person in the group of L. interrogans reported a history of fever in the year of follow-up Supplemental Table 5).

In addition, 8.70% (2/23) of seroincident cases by L. interrogans serogroups reported corn cultivation around their houses. Cassava cultivations were also reported in the peridomiciliary area of seroincident cases for L. interrogans (13.04%, 3/23) and other Leptospira species serogroups (17.65%, 3/17). Households of seroincident cases for both outcomes had a high frequency of rat infestation with 91.30% (21/23) for L. interrogans and 76.47% (13/17) for other Leptospira species. Only the seroincident cases for other Leptospira species serogroups reported hunting canines (17.65%, 3/17) Supplemental Table 5).

The multivariable model evidenced outdoor occupations in the past year (RR = 2.93; 95% CI: 1.36–6.16), the presence of rats (RR = 2.81; 95% CI: 1.07–7.12), and corn cultivation around the households (RR = 22.74; 95% CI: 5.75–51.74) as risk factors for the seroincident cases of L. interrogans serogroups. Furthermore, age in years (as a quantitative variable) was a risk marker (RR = 1.02; 95% CI: 1.01–1.04), and male gender was a protector marker (RR = 0.34; 95% CI: 0.15–0.80) for the seroincident cases of L. interrogans serogroups (Table 4). Likewise, the multivariable model for seroincident cases of other Leptospira species serogroups adjusted by occupation, age, and gender showed that the presence of hunting canines (RR = 6.17; 95% CI: 2.72–10.99) and cassava cultivation (RR = 3.05; 95% CI: 1.24–6.52) were risk factors (Table 4). We did not find any effect modification between occupation and gender in the binary and multinomial outcomes.

Table 4.

Multivariable analysis of seroincident cases of Leptospira among different serogroups (mixed-effects multinomial logistic regression model)

	L. interrogans serogroup	Others Leptospira species serogroups
Variables	RRadjust (95% CI)	RRadjust (95% CI)
Individuals
Outdoor occupation*	2.93 (1.36–6.16)	2.02 (0.80–4.63)
Male gender	0.34 (0.15–0.80)	0.50 (0.19–1.30)
Age (years)	1.02 (1.01–1.04)	0.99 (0.98–1.01)
Peri domiciliary area characteristics
Vegetation
Presence of corn culture*	22.74 (5.75–51.74)
Presence of cassava culture*		3.05 (1.24–6.52)
Synanthropic mammals
Presence of rats*	2.81 (1.07–7.12)
Animal husbandry purpose
Presence of hunting canines*		6.17 (2.72–10.99)

^*The reference category is the absence of the characteristic evaluated.

Among L. interrogans seroincident cases, the most frequent serogroup seropositivity was against Bataviae, followed by Djasiman and mix serogroups (Figure 3). Of note, 100% of seroincident cases of Djasiman, Icterohaemorrhagiae, and Autumnalis serogroups were exposed to rodents in their households, and 88.88% of seroincident cases to Bataviae serogroup had the same exposure to rodents. Among seroincident cases to other Leptospira species serogroups, the most frequent seropositivity was against Tarassovi, followed by Panama (Figure 3). Furthermore, 50% (3/6) and 60% (3/5) of seroincident cases to Tarassovi and Panama serogroups had pigs and opossums around their households, respectively.

Figure 3.

Frequency of Leptospira interrogans and other Leptospira serogroups in the seroincident cases.

DISCUSSION

Leptospirosis is a zoonotic disease with mandatory notification to the National Surveillance System in Public Health in Colombia since 2007. However, the program fails to detect active cases of leptospirosis in rural areas due to the low access to diagnostic tests and the lack of medical care.²⁰ Thus, conducting prospective studies to analyze and detect potential areas at risk of transmission of tropical diseases and estimating risk and protective factors for subsequent interventions are becoming increasingly necessary. The Urabá region in the Department of Antioquia, Colombia, is considered a potential endemic region for tropical diseases such as malaria, leishmaniasis, dengue viruses, leptospirosis, and rickettsiosis, among others.^15,28,29 Therefore, analyzing seropositivity outcomes for more than one disease when conducting analytical observational studies could broaden the understanding of factors for those tropical and neglected diseases potentially endemic in this region.

The overall Leptospira seroprevalence estimated in the current study was 27.81%, very similar to the report from San Juan, Puerto Rico (27.2%, 55/202).³⁰ Additional studies conducted in the Urabá region using a definition of seropositive outcome ≥ 1:100 titers by MAT, reported a Leptospira seroprevalence of 35.6%¹⁴ and 12.5%.¹³ We estimated an overall cumulative seroincidence of Leptospira of 14.60%; to the best of our knowledge, this could be the first estimation of Leptospira seroincidence in Colombia, which helps to identify potential groups at risk of leptospires transmission.

In the seroprevalence analysis, male gender was a factor associated with Leptospira infection in both binomial and multinomial analysis. In fact, the stratified analysis by gender showed that outdoor occupation was a risk factor for Leptospira infection only in males. Outdoor occupation was also a factor associated with seroincident cases of Leptospira infection (multinomial and binary outcomes), as well as older age. This exposure (outdoor labor) is recognized as an occupational risk for Leptospira infection.^3,25 These results are confirmed by our demographic data, showing that male individuals over 15 years old likely work outdoors in the Urabá region. Furthermore, the risk markers identified in our study, such as male gender and age, were also reported as factors associated with Leptospira infection in previous studies.^6,31 Strikingly, our results showed that male gender was a protective factor in the seroincidence analysis. This unexpected result could be the consequence of a large number of male participants lost to follow-up (potential selection bias). Another potential explanation is that women in rural areas of Urabá region are more exposed than men to L. interrogans serogroups because they remained in the intra- and peridomicilliary area most of the time, where they are more likely to be in contact with some infection sources, such as urine from rats and other synantropic or domestic mammals. In fact, we found that women had a higher risk of seropositivity to L. interrogans serogroups when exposed to piglets in the peridomiciliary area. This finding should be confirmed in future studies.

Other characteristics found in our study area, such as poverty and some cultural practices (e.g., an abundance of domestic animals and precariousness in household construction) are conditions highly associated with the presence of a large number of tropical zoonotic diseases.³² In our study, the presence of dirt floors and opossums in peridomiciliary areas were factors associated to seroprevalent cases of L. interrogans serogroups. Other studies have identified the presence of Leptospira infection in pigs and opossums in the peridomiciliary area; these mammals have been considered as potential hosts participating in the life cycles of several Leptospira species.^3,33,34 Bierke et al.³⁵ found that Leptospira can survive for long periods in several types of soil depending on its physical, chemical, and biological conditions such as pH, temperature, or mineral concentration. Furthermore, a significant amount of Leptospira has been detected in 31% (22/70) of samples from different soil microenvironments in a neighborhood endemic for leptospirosis in Salvador, Bahia, Brazil.³⁶

The presence of pit latrines around households and turkeys as domestic animals were risk factors for seropositivity against other Leptospira species serogroups. Similarly, a lack of toilets inside the household was a factor associated with Leptospira seropositivity (unadjusted OR = 3.75; 95% CI: 1.33–10.60) in a study carried out in southern Colombia.³⁷ Recently, a study in Germany identified the presence of poultry in peridomiciliary areas as a risk factor for Leptospira infection.²⁵ However, it is most likely that the presence of poultry or turkeys in peridomiciliary areas is associated with the presence of synanthropic rodents or wild mammals, a necessary condition to maintain leptospires in the environment.³⁸

Likewise, the presence of rodents in intra- or peridomiciliary areas and the presence of corn cultivation in peridomiciliary areas (proxy of the presence of synanthropic and wild mammals), were factors associated with seroincident cases of L. interrogans serogroups, whereas the presence of hunting canines and cassava cultivation in peridomiciliary areas were risk factors for seroincident cases for other Leptospira species serogroups. Similar to urban areas of Brazil, where only the L. interrogans serogroup is circulating, household conditions (household income), sociodemographic aspects (age, gender, and occupation), and rat infestation were risk factors for Leptospira infection.^5,6,31,39 Regarding canines, these domestic animals have been considered important reservoirs or hosts in the infection cycle of Leptospira because they can expel the leptospires through their urine for long periods of time (weeks or months).⁴⁰ Different serogroups of Leptospira, such as Canicola, Autumnalis, Australis, Grippotyphosa, Harjo, Pomona, Pyrogenes, and Djasiman, among others,^39,41–43 have been found in canines. In Brazil, a longitudinal study of dogs found a seroprevalence between 9.3% (95% CI: 6.7–12.6) and 19% (95% CI: 14.1–25.2) in different periods. Additionally, an overall cumulative seroincidence of 11% (95% CI: 9.1–13.2) was estimated, and the cumulative seroincidence by trimester varied between 6% (95% CI: 3.3–10.6) and 15.3% (95% CI: 10.8–21.2). In particular, unvaccinated pig-hunting dogs also have been found seropositive to Leptospira spp. (mainly Australis serovar) in Queensland (23/87; 26%).⁴⁴ A study in Brazil found 10.6% (16/170) of hunting dogs were seropositive for at least one Leptospira spp. serovar.⁴⁵ These domestic animals could be sentinels of Leptospira infection in rural areas, inclusive may play role in the transmission of leptospires from wild animals to human and domestic animals.

Leptospira infection in the Department of Antioquia is most commonly caused by serogroups of the L. interrogans species.⁹ A study conducted in the Urabá region on patients with leptospirosis showed a high proportion of infection for L. interrogans serogroups of Copenhageni (28.6%), Hardjo (20.6%), Australis (17.5%), Canicola (11.1%), and Pomona (6.3%). Other Leptospira species serogroups have a high infection proportion, including Tarassovi (25.4%), Grippotyphosa (17.5%), Shermani (9.5%), Ballum (7.9%), Cynopteri (3.2%), and Panama (1.6%).³² Moreover, in the same study, of 199 patients with febrile syndrome, 11.56% were infected by L. interrogans serogroup Icterohaemorrhagiae, 10.05% were infected by Sejroe, and other serogroups registered a proportion of infection near to 10%.²⁰ Regarding seroprevalent cases in the Urabá region, it was reported that 87% of seroprevalent cases had high titers against Icterohaemorrhagiae, followed by the Sejroe serogroup.¹⁴ Likewise, in a study conducted in the city of Cali in the south of Colombia, the seroprevalence of Leptospira was 12.53%, and the most frequent serogroup in seroprevalent cases was Australis (61.4% of prevalent cases).³⁴ In our study, seroprevalent and seroincoident cases were infected mainly by L. interrogans serogroups Bataviae, Djasiman, and Icteroharmorrhagiae, and the most common other Leptospira serogroups were Tarassovi, Panama, Cynopteri, and Grippotyphosa. The differences of the frequency of serogroups found among our study and the previous studies could be driven by local difference in the study area (urban versus rural) and other factors related to the presence of amplifying host, work activities, and closer contact between humans and domestic animals such as pigs, equines, bovines, and poultry in rural areas.

The highest exposure to Bataviae among prevalent and incident cases of L. interrogans seropositive individuals, indicates that these cases had probably been exposed to infected rodents or canines.^46,47 Nevertheless, detection of antibodies against Djasiman and Icterohaemorrhagiae serogroups evidenced additional exposure to sources contaminated with leptospires shed by synanthropic or wild mammals.^3,33 Regarding different species of L. interrogans serogroups, this study showed a high frequency of seroprevalent and seroincident cases against the Tarassovi serogroup, which is mainly found in pigs.^3,34 Even though the Tarassovi serogroup is present in different species of Leptospira considered as pathogens and classified into subgroup II, this serogroup has not been implicated in clinical cases of leptospirosis in humans.⁴⁸ Panama was another important serogroup found in seroprevalent and seroincident cases. This serogroup has been detected in opossums, wild rodents, and numerous domestic animals.^49–52 However, there are several clinical cases of leptospirosis described in humans, potentially caused by the Panama serogroup.^50,53 Finally, reactivity to the Cynopteri serogroup was third in seroprevalent cases and fourth in seroincident cases. This serogroup is frequent in bats and various wild animals.^3,49,54

Although multinomial regression modeling is infrequently used to analyze infectious diseases,⁵⁵ it has been used to study the seroepidemiology of Leptospira infection in cross-sectional and prospective studies.^6,56,57 This method had the advantage of estimating the simultaneous effects of multiple factors. In our study, we analyzed the factors associated with the risk of Leptospira infection, stratified in three categories (L. interrogans serogroups, other Leptospira species serogroups, and seronegative). In contrast to the binary analysis, which only permits comparison of seropositive and seronegative individuals, the multinomial analysis allows a better understanding of factors associated to specific categories of the outcome (L. interrogans and other Leptospira species serogroups). In addition, the inclusion of seroprevalence and seroincidence analyses is useful in studying leptospiral infection because they helped to understand the factors associated to the burden of infection and to the risk of occurrence of new leptospiral infections in the population, respectively.

The main limitation of this study was the moderate sampling coverage of individuals (58.41%). Some participants declined study enrollment, and others moved out in search of work. Consequently, selection bias was likely present in this study at the analysis unit (inhabitants of households). Another important limitation was the information bias resulting from obtaining group information of participants enrolled in each household through the family head, such as cultural practices shared by all household inhabitants. However, this study considered that some habits and perceptions are widespread in households depending on the region or area where people have grown or lived; for instance, the risk perceptions related to infectious diseases, self-care measures, or land use. For the seroincidence analysis, the main limitation was the losses to follow-up. Selection bias was likely introduced in the estimation of the overall cumulative seroincidence, serogroups seroincidence, and associated factors because the majority of the losses were males. Another potential limitation was the sensitivity and specificity of the microagglutination test to detect the seroincident and seroprevalent cases, especially in seropositive individuals who had titers equal to 1:50. Nevertheless, the sensitivity and specificity of the test could be increased when diverse serovars are included, especially those circulating in the same area. In this study, 32 Leptospira strains representing several species and serogroups (Table 1) were included, and a random selection of 20% of the total human serum samples collected at baseline was used to select the final panel of 16 Leptospira strains tested in this study. In addition, the classification of most probable infecting serogroup (L. interrogans and other Leptospira species serogroups) by MAT is not perfect and it could result in misclassification bias. Another limitation is that we found sparse data in some variables, which can increase type I error. Finally, information biases are a concern in the followed-up participants—specifically, memory biases respecting changing exposures such as episodes of fever; sporadic outdoor occupation; and contact with synanthropic, wild, or domestic mammals, among other variables.

CONCLUSIONS

The current study aimed to understand the potential public health problem of Leptospira transmission in the rural areas of Urabá, Colombia, where fatal cases of leptospirosis have occurred. The estimated Leptospira seroprevalence was 27.81%, with infections being predominantly caused by serogroups of L. interrogans (65.66%). The overall seroincidence of Leptospira was 14.60%, and a contrast between seroprevalence results with a similar distribution between L. interrogans and other Leptospira species serogroups was evident. We identified important risk factors, including different activities and characteristics related to Leptospira infection caused by different species or serogroups of Leptospira in the area. The sequential binary and multinomial regression analysis of this study allowed us to obtain the estimated risk and protective factors associated with Leptospira infection (seroprevalent and seroincident cases) in rural areas of the Urabá region. The characteristics of the cases of Leptospira infection identified in this study agreed with the cases reported in the Colombian surveillance system for Leptospira.⁵⁸ This rural area seems to have different species of Leptospira circulating than other areas, such as Salvador Bahia, Brazil, which is a more urban area and where synantropic rodents have a major role in transmission.⁵⁹ However, this study is consistent with other works showing that transmission in rural areas is also dependent on wild animals and environmental conditions favoring the risk of infection, making disease control difficult and highlighting the importance of this study, which identified factors associated to Leptospira infection. The long-term goal of this work is to strengthen the surveillance system by identifying groups at risk for Leptospira transmission focusing the potential intervention and control measures in the affected communities at the Uraba region. This, in turn, will aid targeted efforts to provide appropriate diagnosis and timely treatment and ultimately reduce the case fatality rate associated with the disease. Finally, this study will help to prioritize and develop strategies to prevent incident cases in this area and will help to strengthen educational strategies to prevent the transmission of the disease in the communities from rural areas of Colombia.

Note: Supplemental materials appear at www.ajtmh.org.