Prevalence and risk factors for intestinal parasitic infections in pregnant women residing in three districts of Bogotá, Colombia

A cross-sectional, community-based study, in which participants answered a questionnaire and provided one or two stool samples, was conducted between May 2015 and July 2016 in Bogotá. This study focused on pregnant women living in poor residential areas in three districts of Bogotá (Fig. 1). These communities were selected because they have a majority of strata 1 and 2 neighborhoods or receive a large number of people displaced within Colombia [50], who live in socioeconomic and geographical conditions that may affect their risk of IPI. The selected districts were Usaquén (population 472,908; area 65.31 km²), Kennedy (population 1,187,315; area 38.56 km²), and Ciudad Bolívar (population 719,700; area 129.98 km²) [42, 51].

Colombia’s health care system is organized into three levels of attention with primary care units (Unidad Primaria de Atención or UPA) within a basic level of health care and secondary and tertiary care hospitals. In Bogotá, antenatal care in public health settings is provided in UPAs with high-risk cases being referred. The study’s target population was all pregnant women in any trimester, living in the districts of Usaquén, Kennedy, and Ciudad Bolívar, belonging to strata 1 or 2, and attending antenatal care in local primary care units. During the study period, a total of about 9600 pregnant women attended these units.

To calculate the sample size in this study, the Epi Info™ 7.2 software developed by the Center for Disease Control and Prevention (CDC, Atlanta, GA, USA) was used. More specifically, using the StatCalc module for cross-sectional studies, sample size was estimated based on exposure and outcome prevalence reported in previous similar studies in Colombian and Venezuelan cohorts. These reported 43%, 12%, and 9.5% parasitism in groups exposed to unfavorable socioeconomic and sanitary conditions, in contrast to 33% and estimated 6% and 4.5% in unexposed groups [22, 24], with estimated ORs of 1.53, 2.14, and 2.23 respectively. The different sample sizes were determined with a level of accuracy of 5% and statistical power of 80%. Thus, sample sizes of 776, 778, and 894 were calculated. Considering a non-response proportion of 20%, the resulting estimated sample size was 1100 pregnant women.

With respect to national data protection regulations, it was not possible to access patient databases directly to establish a sample frame of pregnant women who attended the primary care units within the three districts. Therefore, from May 2015 to July 2016, research assistants proactively invited pregnant women to participate in the study. Fieldwork was supported by six research assistants experienced in community work who had received training in technical, ethical, and logistic procedures before data collection. Once the informed consent form was signed, each woman received a kit and instructions for appropriate stool collection. Sociodemographic aspects and pregnancy characteristics were assessed at the UPAs through individual interviews, while data related to housing conditions and hygiene habits were assessed through questioning complemented with inspection during home visits. Once the questionnaire was completed, pregnant women received an educational intervention about general health recommendations and danger signs during pregnancy. Afterwards, participants received their exam results with a recommendation to include them in the next visit to the physician. Fieldwork started in the districts of Usaquén and Kennedy. In these districts, research assistants approached pregnant women directly before and after prenatal courses at UPAs and, in addition, in Usaquén, they included door-to-door visits. Ciudad Bolívar was the last district, in which research assistants systematically invited every pregnant woman who attended any time between 7 am and 4 pm at two UPAs selected within the district because both received 80% of pregnant women from the district. In the district of Usaquén, recruitment had to be suspended as a result of unsafe working conditions for research assistants when neighborhood conditions deteriorated during the period of fieldwork. This led to an underrepresented 4% (24/550) sample of participants in this district. In the district of Kennedy, the recruitment of participants was stopped because of selective sampling in favor of the pregnant women who could attend courses at the UPA. Consequently, only 23% (127/550) of the study participants were from that district. For these reasons, 73% of the participants were from Ciudad Bolívar, the district of Bogotá where 59% of people live in stratum 1 and 38% in stratum 2 [52].

Of the 210 participants who did not provide stool samples, 95% (199/210) were from the district of Ciudad Bolívar, of whom 84% (167/199) resided in stratum 1 (Table 1). This district is located on a mountain with the poorest homes near the top, where access is difficult and not always feasible for geographical and safety reasons. Despite these limitations in accessing these participants, research assistants attempted data collection when conditions allowed. Additionally, some pregnant women who initially answered the questionnaire in the local primary care units did not accept the home visit to pick up the stool sample and did not hand in the sample to the units.

A total of 775 pregnant women were invited to participate: of these, 71% (550/775) accepted the invitation; of the latter, 94% (519/550) answered the questionnaire completely and 62% (340/550) gave one stool sample. Of the 340 participants who provided a stool sample, 32% (107/340) provided a second stool sample, nine of which had to be excluded because of processing errors. Thus, this study included 331 participants with at least one stool sample, 98 of whom provided a second stool sample. In addition, of the 331 participants, a subgroup of 50 samples was processed further for quantitative polymerase chain reaction (qPCR) investigation. Two of these samples were lacking formol–ether concentration data, so they were excluded from further comparison, resulting in 48 samples included for molecular analysis.

The interviewer-administered questionnaire was adopted from the Demographic and Health Surveys (DHS) to assess sociodemographic conditions as implemented by ICF International (Fairfax, VA, USA) [53]. The items included were age, occupation, education level, civil status, health insurance coverage, monthly income, household conditions, water availability, supply, and sewage, and hygiene habits such as boiling water before drinking, washing fruit and vegetables, place and reasons for washing hands throughout the day, and habits of walking barefoot. Two research members (AE and KR) translated the questionnaire from English into Spanish. In addition, questions about parity and trimester of pregnancy were taken from a Spanish form on “Pregnant women in primary health care” [54]. Items on garbage collection, vectors, and living with pets were chosen from the Spanish form on “Elementary family characteristics” [55], both developed by the Health District Secretary of Bogotá. Supplementary questions regarding the location of the house, socioeconomic stratum, and date of the last deworming procedure were included. A pilot test of the questionnaire to control for clarity, comprehension, and duration of the survey was carried out with three students from Universidad del Rosario, Bogotá. They were in their fifth year of medical school and received training by two investigators (AE and AP) to apply the survey. Ten women residing in the district of Usaquén who were in their last month of pregnancy were interviewed once they gave their informed consent. These women were not included as participants in the study. Minor wording adjustments were made following their feedback.

All study participants received stool sample containers and standard instructions on proper and safe collection and preservation of the samples. By agreement, research assistants contacted participants by phone to insure that, once the sample was collected and stored in proper conditions (inside the fridge or in a dark low-temperature home location), the assistants could pick it up from the participants’ homes no more than 4 h after evacuation. The research assistants, who were trained in biosafety standards, followed a protocol for stool sample collection and handling, including appropriate labeling, triple packaging [56], keeping the cooling chain, and ensuring proper, timely, and safe handling and delivery of stool samples to the laboratory.

The detection techniques selected in this study were based on diagnostic performance, methodological availability, feasibility, and cost-effectiveness [57]. In this study, each stool sample was analyzed by a so-called “combined microscopy technique”, which included sample analysis by direct wet mount microscopy and by a formol–ether concentration method and subsequent microscopy [58]. Both methods were performed by a contracted specialized nationwide reference laboratory following standard procedures. Direct wet mount microscopy has reportedly shown better diagnostic capacity for protozoal trophozoites, especially Giardia [57] and Blastocystis [59], while the concentration technique has demonstrated better performance for other protozoa [59] and for helminths [60, 61].

In this study, a subgroup of 48 stool samples was subsequently analyzed by qPCR for comparison with the results obtained by standard microscopy detection techniques. These samples originated from participants residing in Ciudad Bolívar district and were selected prospectively and chronologically by order of arrival at the laboratory. For financial reasons, not all samples could be analyzed by qPCR. The qPCR technique, with 100% primer-determined specificity and reported sensitivity close to 100% [62, 63], was carried out by the microbiology laboratory of Universidad del Rosario, Bogotá.

At the university laboratory, the stool sample was fixed with 70% ethanol. With 300 μl of the fixed sample, DNA was extracted using the Norgen Stool DNA isolation kit (Norgen Biotek Corporation, Thorold, Canada) following the manufacturer’s protocol. The PCR was conducted on a total volume of 7 μl containing 3.5 μl of Taqman Mastermix™ (Applied Biosystems, Foster City, CA, USA), 2 μl of DNA, and 1.5 μl of species-specific primers (final concentration = 900 mM) and Taqman probes (final concentration = 100 mM). Samples were run in an Applied Biosystems 7500 Fast Real-Time PCR system and processed using a denaturation time of 3 s at 95 °C and an extension time of 30 s at 60 °C for 40 cycles. The results of qPCR were considered negative if the cycle threshold values (Ct) were > 38. This threshold was determined by measuring the detection limits of each assay in which serial dilutions of parasites were included [64]. The available PCR primers were: Blastocystis hominis, Giardia lamblia, Cryptosporidium, Entamoeba histolytica, Ancylostoma duodenalis, Necator americanus, Ascaris lumbricoides, and Trichuris trichiura, all of which were based on the primers reported by Mejia et al. [64], except the primer for B. hominis, which was based on Stensvold et al. [65]. A reference strain of Giardia duodenalis WB and DNA from each of the parasites were used as positive controls. Stool samples previously collected and confirmed by the laboratory to be from healthy non-infested children served as negative controls.

All data were double entered and controlled for errors. Data analysis was performed using IBM SPSS Statistics version 24 software (Armonk, NY, USA). A comparison of sociodemographic characteristics was performed between the 210 pregnant women who answered the questionnaire only and the 340 participants who also provided a stool sample.

Laboratory reports indicated the presence of any parasite form (trophozoites, cysts, eggs, or larvae) as detected by direct and concentration techniques separately and for each parasite. With these data, one variable for each parasite was defined as “negative = 0” (if no form was reported) or “positive = 1” (if any parasite form was reported). Once the prevalence of each parasite was established with each technique, a pooled analysis with both techniques was generated. A combined variable for each parasite was created and defined as “negative = 0” (if no parasite was detected by either technique) or as “positive = 1” (if any parasite was detected by either technique). In all included participants, the prevalence was analyzed in three dimensions, namely the presence of any parasite, the presence of any pathogenic parasite, and the presence of more than one parasite (“polyparasitism”). For each outcome dimension, one compound variable was created and defined as “negative = 0” (if no parasite forms were detected with the combined microscopy technique) or “positive = 1” (if parasite forms were detected with the combined microscopy technique).

When qPCR was added to the combined microscopy technique in a subgroup of participants, the prevalence of each parasite was established separately and combined by creating variables, as described above. The percentage of agreement between the two techniques was then determined. Positive agreement percentage was estimated as a proportion corresponding to the number of parasites detected by combined microscopy technique and confirmed by qPCR over the total number of parasites detected by qPCR. Negative agreement was estimated as a proportion corresponding to the number of stool samples reported as negative for parasites by combined microscopy technique and confirmed by qPCR over the total negative cases detected by qPCR [66].

Exposure variables such as sociodemographic and pregnancy characteristics, living conditions, and hygiene habits were evaluated as categorical variables and presented as absolute and relative frequencies. Age was evaluated numerically. Bivariate analysis was done using Pearson’s Chi-square (χ²) test, Fisher’s exact test, or Mann–Whitney test as appropriate. Factors that correlated highly with the polyparasitism variable (p < 0.1) were included in a logistic regression analysis.

Of the 550 pregnant women who agreed to participate, 38% (210/550) only answered the questionnaire and 62% (340/550) responded to the questionnaire and handed in one stool sample. Sociodemographic conditions based on participant characteristics were compared between participants only responding to the questionnaire and those additionally providing a stool sample. Statistically significant differences were identified according to district, stratum, and health insurance coverage between participants who only answered the questionnaire and those who also provided a stool sample (p < 0.001). Among the women who answered the questionnaire only, 94.8% (199/210) resided in Ciudad Bolívar, 80% (168/210) lived in stratum 1, and 80% (168/210) were covered by social health insurance (Table 1). In addition, the median age for participants who only answered the questionnaire was 21 years (range 15–41 years), while the median age for those who also provided a stool sample was 22 years (range 14–43 years), p = 0.034. All other variables showed no statistically significant differences.

Nine participants from the 340 (3%) who delivered at least one stool sample had to be excluded because of processing errors in the laboratory. Of the remaining 331 participants, 107 (32%) voluntarily provided a second stool sample, and nine of the 107 (8%) had to be excluded because of processing errors in the laboratory. The 331 participants with at least one analyzed stool sample will serve as denominator in the following analyses. Among those who provided one stool sample, 41% (CI 95% 35.7–46.3) (137/331) had at least one intestinal parasite, either pathogenic or non-pathogenic, and 9% (CI 95% 5.9–12.0) (31/331) had more than one intestinal parasite (polyparasitism). The overall prevalence of any pathogenic parasites was 1.2% (CI 95% 0.0–2.4) (4/331), consisting of two parasite species. As shown in Fig. 2, the prevalence of Giardia lamblia was 0.9% (3/331) and of Ascaris lumbricoides was 0.3% (1/331). The findings for the Entamoeba histolytica/dispar complex have to be regarded separately, as this parasite was only detected in those samples that were investigated by microscopy-based methods; hence, a differentiation between Entamoeba histolytica as a pathogenic species and Entamoeba dispar as a non-pathogenic species, which is only possible by nucleic amplification techniques, was not available. The prevalence of the complex was 1.5% (5/331). Regarding non-pathogenic parasites, the prevalence was 25% (83/331) for Blastocystis hominis, 15% (50/331) for Endolimax nana, 8% (26/331) for Entamoeba coli, and 2% (6/331) for Iodamoeba butschlii.

Of the 98 participants who provided a second stool sample, when both the first and second samples were analyzed together with combined microscopy techniques, the prevalence of any parasite was 52% (51/98), with 14% (14/98) for more than one parasite. The only pathogenic intestinal parasite detected was Giardia lamblia in 3% of the samples (3/98). For non-pathogenic parasites, the prevalence of Blastocystis hominis was 36% (35/98), of Endolimax nana was 22% (22/98), of Entamoeba coli was 9% (9/98), and of Iodamoeba butschlii was 3% (3/98). In addition, the prevalence of Entamoeba histolytica/dispar complex was 2% (2/98).

As seen in Fig. 3, the second stool sample increased detection from the first to the second stool sample from 37% (33/98) to 52% (51/98) for any parasite and from 9% (9/98) to 14% (14/98) for polyparasitism, while it remained unchanged for pathogenic parasites at 2% (2/98).

Test results for a subset of 48 samples were compared between qPCR results for eight selected parasites, as outlined in the methods section, and the combined microscopy-based techniques. Two of these 50 samples could not be conclusively processed in the combined microscopy technique and therefore had to be excluded from this sub-analysis. For the eight investigated parasites, qPCR identified only two types of parasites, B. hominis in 54% (26/48) and G. lamblia in 4% (2/48). The combined microscopy technique identified B. hominis in 27% (13/48) and 0% (0/48) for G. lamblia. When comparing both techniques to identify the presence of any parasite in the samples, the prevalence estimated by qPCR was 54% (26/48) in contrast to 31% (15/48) with the combined microscopy technique (p Fisher < 0.004). The positive and negative agreements to diagnose any parasite were 50% (13/26) and 91% (20/22), respectively, while positive and negative agreements to diagnose B. hominis were 48% (12/25) and 96% (22/23) respectively. With G. lamblia, there was a positive agreement of 0% and a negative agreement of 96% (Table 2).

Associations of infection by any parasite and intestinal polyparasitism with sociodemographic characteristics (Table 3), living conditions, and hygiene behaviors (Table 4) were tested. A higher significant association was found between infection by any parasite and last deworming: less than 1 year ago with 34% (20/59) prevalence, more than 1 year ago with 52% (55/106) prevalence, and never with 36% prevalence (48/135) (p = 0.01). For infection by any parasite, a higher but non-significant association was found with civil status: being single, 34% (34/101) prevalence; and living with a partner, 45% (103/230) prevalence (p = 0.06). There was no statistically significant difference for the variable of age between women with any parasite infection (median age 21 years (range 14–43 years)) and women without any parasite infection (median age 23 years (range 14–40 years)) (p = 0.16). Higher but non-significant associations with polyparasitism were found for women from minority groups with 21% (5/24) prevalence compared with 9% (26/306) for women from majority groups (p = 0.06), and for women without a water sink in the toilet with 14% (14/100) prevalence compared with 7% (17/230) for those possessing a water sink (p = 0.06). There was no statistically significant difference for the variable of age between women with polyparasitism (median age 22 years (range 17–37 years)) and women without polyparasitism (median age 22 years (range 14–33 years)) (p = 0.09). Similarly, there were no significant differences when polyparasitism was compared between parous women at 12% (18/149) and nulliparous women at 7% (13/181) (p = 0.09). A multivariate analysis was performed including variables in which associations in bivariate analysis resulted in p values lower than 0.10. For infection by any parasite, the analyzed variables were civil status and last deworming. The only variable that remained in the equation was last deworming (more than 1 year p = 0.82, OR 0.929, CI 0.488–1.770; never dewormed p = 0.011, OR 1.955, CI 1.163–3.284). For polyparasitism, the analyzed variables were ethnicity, place to wash hands, age, and parity. The only variable that remained in the equation was ethnicity (p = 0.057, OR 2.82, CI 0.971–8.154).