Risk factors for asymptomatic malaria infections from seasonal cross-sectional surveys along the China–Myanmar border

The study protocol was reviewed and approved by institutional review boards of China Medical University, China, Pennsylvania State University, USA and the local Health Bureau of Kachin State, Myanmar. Before carrying out the study, all adult participants or legal guardians of children voluntarily signed the informed consent.

This study was conducted in 4 villages (MLH, L1, MSY, JHK) and 2 IDP camps (N2 and N3) near Laiza township (97.56°E and 24.75°N), Kachin State, Myanmar, along the China–Myanmar border area (Fig. 1). The study sites are located in the same valley area with an elevation ranging from 200 to 550 m above sea level. In each of the IDP camps, there is a health clinic providing routine health services to the camp residents, while the 4 villages are served by a clinic located in Laiza town. This region has a sub-tropical climate with November to April the dry and cold months, and May to October the warm and humid months. Malaria transmission in this region is perennial and dynamic, concentrated in the rainy season [33]. Anopheles minimus s.l. is the predominant vector species complex, and adult Anopheles density on average is higher in villages than IDP camps [32].

Sample size required to detect the prevalence of asymptomatic infection in the survey sites was calculated as n = (μ²)(π)(1 − π)/δ², using an expected asymptomatic malaria prevalence for the border (π) of 20% [8]; μ statistic of 1.96 for 95% confidence interval (CI) and marginal error (δ) of 2%. Three cross-sectional surveys were conducted in 2015 to assess malaria prevalence in different seasons: March and April, July and November to represent the pre-peak (early), peak, and post-peak (late) seasons, respectively. To encourage participation in this study, a team comprised of nurses from the local Health Bureau visited the villages and camps to inform village leaders and camp authorities about the study. During the surveys, trained local nurses and medical technologists visited the households in the villages and IDP camps to obtain informed consent from the head of each household prior to the interview and sample collection. Instructed questionnaires were used to collect individual demographic and behavioural (gender, age, occupation, malaria history, bed net usage) and household information (education of the household head, major building material of the house, its elevation and spatial location relative to landmarks such as road, forest, water body, border, and clinic, as well as the use of indoor residual spray (IRS). Age was arbitrarily grouped into 2 categories: ≤ 15 and > 15 years, whereas occupation was grouped into 4 categories: students (pre-school and school children), office and in-house workers (office staff, teachers and housewives), soldiers, and farmers. The information on characteristics of households was obtained mainly from heads of household. Housewives were respondents if the heads were away at work. The distance to the indicated locations was calculated in ArcGIS 10.2. Major building material of the house was used as a proxy for the economic status of the family, as economically better families would invest on improvement of house structure with bricks in contrast to the poor living in houses or huts made of naturally available, cheaper materials such as wood or bamboo [34]. During the visit, axillary temperature was measured and those with temperature ≥ 37.5 °C or a history of fever during the prior 24 h were referred to the nearest local health staff or clinic to receive adequate treatment and follow-up as clinically appropriate. Finger-prick blood was used to make thick and thin smears, which were air dried, transported to a nearby field laboratory, and stained with 10% Giemsa. In addition, 100 μL of finger-prick blood were collected in tubes containing EDTA and transferred on ice to the laboratory, where it was mixed with 200 μL Trizol reagent and stored at − 80 °C. The samples were shipped on dry ice to China Medical University for laboratory studies.

Thick and thin blood films were screened for the presence of malaria parasites using light microscopes with 100× oil immersion lens. Two microscopists examined all slides, and when discrepancy occurred, a third microscopist read the slides and the results were combined. For RNA extraction, 1 mL of Trizol reagent was added to the frozen sample containing 0.1 mL of blood. Total RNA was isolated according to the manufacturer’s protocol and dissolved in 30 μL RNase-free water. One microgram of each RNA sample was used to synthesize cDNA using the Takara RNA PCR kit (AMV) version 3.0 (Takara, Japan). Parasite detection was performed by nested RT-PCR (nRT-PCR) targeting the 18S rRNA using an established protocol [8], which showed a limit of detection of 0.01 parasites/μL of blood.

Maps were generated using Google Earth. Statistical analysis was performed using SPSS Statistics 19.0. For risk factor analysis, both univariate and multivariate mixed effects logistic regression models were applied. First, explanatory variables were tested individually in univariate analysis. Then, all factors with P < 0.25 were selected to be involved in the following multivariate analysis [35, 36]. Potential risk factors were analysed at both individual and household levels, and only those with P < 0.05 in multivariate analysis were considered to be the risk factors. Individual-level, independent variables included gender, age, occupation, malaria history, and bed net usage. Household-level factors included education of household head, building material of the house, IRS usage and location elevation. In addition, spatial location relative to landmarks was acquired for each household and considered household-level variables for analysis. For descriptive analyses, χ² test or Fisher’s exact test was used. Asymptomatic malaria prevalence for the predictive factors was compared between the rainy and dry seasons, and between villages and IDP camps using χ² test, and odds ratio (OR) and adjusted odds ratio (AOR) were calculated. P < 0.05 was considered statistically significant.

A census prior to the surveys identified the population size in the 4 study villages as 1436. The IDP camps were established in 2011 as the result of Myanmar civil war with an initial population of ~ 7000, which was increased to 10,402 in 2015 with populations migrated from more interior areas of Kachin. Using the prevalence of ~ 20% of asymptomatic infections determined using a sensitive molecular method [8], the sample size to detect malaria parasite infection prevalence was determined to be 1537. Thus, the sample size was expanded to include 1680 healthy individuals without fever and malaria-related symptoms from 773 households, which represented 38.4% (773/2014) of the total households (Fig. 2). Three cross-sectional surveys were conducted in early (n = 516), peak (n = 578) and late (n = 586) seasons. Though the census identified 48.8% of the population as males, many adult males could not be reached as they were required to be enlisted. Therefore, the surveyed population was female-biased (70.3%). The median age of the volunteers was 33 years (95% CI 31–34), with more than 70% participants in > 15 age groups in each season. The majority of the participants were office and in-house workers (54.6%), followed by students (28.3%). When asked about their malaria histories, 3.6% of the participants self-admitted that they had malaria within the last year. With regard to preventive measures targeting vectors, participants were specifically asked for the use of bed nets. The earlier study established that bed net ownership in the study community was close to 100%, but usage was ~ 75%. The current surveys identified that less than half (45.1%) of the participants used bed nets in previous nights (Table 1). Of the 773 households surveyed, 35.4% had IRS during the previous month. The majority of the heads of the households had primary (34.9%) or middle school (63.0%) levels of education. With regard to the house structures and building materials, the majority of the houses were built using wood and bamboo as the main construction materials (95.2%), whereas only 5.2% of the houses were constructed using bricks (Table 2).

Microscopic examination of the 1680 blood smears only detected P. vivax infections in 25 samples, giving an overall 1.5% parasite prevalence. The prevalence of P. vivax infections in the peak season (2.3%) was slightly higher than that of the early season (1.7%) and late season (1.5%), but the differences were not statistically significant (Table 3). In comparison, nRT-PCR targeting the 18S rRNA gene detected 23.4% (393/1680) prevalence of asymptomatic infections. Specifically, 23 of the 25 microscopy-positive samples were confirmed by nRT-PCR, demonstrating that the majority (94.1%) of the asymptomatic infections were sub-microscopic. The 393 nRT-PCR positive samples included 352 (89.6%) Plasmodium vivax, 32 (8.1%) Plasmodium falciparum, and 9 (2.3%) mixed P. vivax/P. falciparum infections. Plasmodium ovale, Plasmodium malariae and Plasmodium knowlesi infections were not detected. It is noteworthy that P. falciparum infections were only identified in the 2 later surveys.

In contrast to the results from microscopic examination, there were significantly seasonal differences in the prevalence of infections detected by nRT-PCR (P = 0.001): 18.4% of infection occurred in early season (March–April), 30.1% in the rainy season (July), and 21.2% in the late season (November) (Table 3). This seasonal variation was notably dramatic among the villagers. Prior to the rainy season, the prevalence of asymptomatic infection in the villages was relatively low at 8.43%, which sharply rose to > 50% during the peak transmission season (Fig. 3). In comparison, the prevalence of sub-microscopic infections did not show substantial seasonal variation in the IDP camps (Fig. 3). Furthermore, the comparison of the asymptomatic infections between villages and IDP camps during the different seasons revealed that before the rainy season, the prevalence of asymptomatic infection in the villages (8.43%) was significantly lower than that in the IDP camps (8.43 vs 23.7%, P < 0.001), whereas this trend was obviously reversed in the peak transmission season (54.1 vs 24.5%, P < 0.001). This trend was maintained after the rainy season, and the prevalence of asymptomatic infections in the villages (26.3%) was much higher than that in the IDP camps (19.3%) in November, though the difference was marginal (P = 0.069).

Univariate and multivariate analyses were performed to identify the risk factors associated with asymptomatic malaria infections. In the univariate analysis, all potential factors were tested and only those with P values less than 0.25 (including gender, age, bed net use, major building material of house, IRS use, elevation, location to clinic and road) were selected and set into the final multivariate model. First, the results showed that males were significantly more likely than females to have asymptomatic infections (AOR: 1.5; 95% CI 1.16–1.97, P = 0.002). Second, the age group of ≤ 15 had significantly higher odds of malaria infection (AOR = 1.49, 95% CI = 1.13–1.96, P = 0.004). Third, the effects of vector control-based preventive measures were very remarkable. Those who did not sleep under the bed nets had more than 5 times higher odds of asymptomatic infections than those who did (AOR: 5.49, 95% CI 4.12–7.32, P < 0.001). People from the households that did not have IRS had 3.4 times higher odds of asymptomatic infection than those with IRS (95% CI 2.50–4.65, P < 0.001). Fourth, major building materials of the houses, which reflect the economic status of the family, also showed significant effect on asymptomatic infections. Residents from the houses constructed from wood or bamboo had approximately twice the odds of having Plasmodium infections (AOR: 1.90, 95% CI 1.15–3.14, P = 0.012). Last, longer distance of household to the nearest clinic (> 180 m) was also significantly associated with a higher risk of malaria infection (AOR: 2.05, 95% CI 1.53–2.75, P < 0.001) (Table 4).

To further explore whether the risk factors for asymptomatic malaria infections experience seasonal variations, the 2 surveys conducted in the dry season (March–April and November) were combined and compared with that conducted in the rainy season (July). With regard to preventive measures, those who did not use bed nets had 10.53 (95% CI 6.38–17.40; P < 0.001) and 3.81 times (95% CI 2.67–5.43, P < 0.001) higher odds of having asymptomatic infections than those who used bed nets in the rainy and dry season, respectively (Table 5). Moreover, those who did use bed nets in the rainy season had 2.28 times higher odds of having asymptomatic infections than those in the dry season (95% CI 1.72–3.03, P < 0.001). Similarly, those without the use of IRS had 7.78 (95% CI 4.57–13.25, P < 0.001) and 2.29 (95% CI 1.54–3.40, P < 0.001) times higher odds of infection in the rainy and dry season, respectively (Table 5), giving a rainy season/dry season risk ratio of 2.43 (95% CI 1.86–3.17, P < 0.001). Furthermore, larger distance (> 180 m) of the household to the nearest clinic was significantly associated with a higher risk of malaria infections in both seasons. For other factors, there were also seasonal variations. Whereas the male gender had increased odds of asymptomatic infections in the rainy season, the factor did not seem to contribute to higher infection prevalence in the dry season (Table 5). Also, people who were ≤ 15 years and lived in houses constructed with wood/bamboo, which were located in lower elevation (≤ 312 m above sea level) and closer to road (≤ 15 m) had higher odds of having asymptomatic infections only in the dry season.

Further comparison of the risk factors for asymptomatic malaria infections revealed additional differences between the villages and IDP camps (Table 6). Although people who did not use bed nets from villages and camps all had increased odds of having asymptomatic Plasmodium infections compared to bed net users, those non-users from the camps had much higher odds of asymptomatic infections (AOR: 7.37, 95% CI 5.15–10.53, P < 0.001) than the non-users from the villages (AOR: 2.65, 95% CI 1.61–4.35, P < 0.001). Similarly, IRS use also significantly reduced the odds of asymptomatic infections in households from both villages and camps, but the effect appeared more profound in the villages (Table 6). Also, larger distance (> 180 m) of the houses to the nearest clinics was significantly associated with higher risks of malaria infections both in villages and camps. There were also village- or camp-specific risk factors. For example, lower house elevation (≤ 312 m above sea level) was significantly associated with higher odds of asymptomatic malaria infection in villages, whereas there were obvious associations of the male gender and the ≤ 15 age group with higher risks of asymptomatic infections in IDP camps.