Temporal Dynamics of Subclinical Malaria in Different Transmission Zones of Myanmar

Study design, sites, and populations.

A prospective cohort study was performed to characterize short-term changes in subclinical infection status among participants living in three epidemiologically distinct regions of Myanmar. The study sites were chosen to represent a range of malaria epidemiologies characteristic of the region, including three military bases and three villages in Monywa Township in Myanmar’s Sagaing Region (referred to henceforth as “Monywa” site); two villages and one refugee community in Laiza Township in Kachin State, located on Myanmar’s border with China’s Yunnan Province (“Laiza” site); and five villages in Ann Township in Myanmar’s Rakhine State (“Ann” site) (Figure 1). Based on surveillance data and case reporting, Ann and Laiza Townships have been characterized as highly endemic with P. falciparum and P. vivax as the predominant species, respectively. Sagaing Region has historically had low malaria endemicity overall but with several “hot-spot” areas, including Monywa Township where both P. falciparum and P. vivax were suspected to be highly endemic. Selection of study villages or communities within each study site was based on known clinical malaria burden (unpublished data from annual reports of the Myanmar National Malaria Control Program and the National Institute of Parasitic Diseases, Chinese Center for Disease Control and Prevention), and on logistical considerations intended to ensure the integrity of study data and samples.

View Full Size

Map of study areas.

Citation: The American Journal of Tropical Medicine and Hygiene 107, 3; 10.4269/ajtmh.22-0027

• Download Figure
• Download figure as PowerPoint slide

Two-step simple randomization was performed to select households within each study village or community, and participants within each household. A nearest next household was selected if originally selected household members were unavailable or declined to participate. Information about the study was disseminated within each study site by community engagement activities and by word-of-mouth, the most effective ways of research study recruitment in these remote communities, based on our previous experience. Inclusion criteria included age at least 6 months old, provision of a written informed consent or assent as appropriate, and ability and willingness to comply with the study protocol. Exclusion criteria included any clinical presentations that required immediate medical attention.

Data collection.

After informed consent was obtained, demographic, medical and malaria history, behavioral, and travel-/mobility-related information were collected following a standardized questionnaire, using Open Data Kit (ODK) software, and uploaded to an Ona server and stored in a secure database at Duke University. Potential risk factors for malaria infection collected at baseline and during the follow-up period were age, gender, date of study data and sample collection (peak season is June–December, non–peak season is January–May), residence (defined as living in the study village for at least 6 months), type of occupation (dependent, student, soldier, refugee, farmer, plantation worker, or other, where “dependent” refers to a participant who did not work), setting of occupation (predominantly indoors versus predominantly outdoors), the presence of seasonal variation in primary occupation, and regular use of a bed net, defined as at least four+ nights per week. Travel patterns that were potentially related to malaria exposure were also investigated, including the distance from home to place of work or study, greatest distance traveled in a typical day, frequency of travel outside of the village in the last 6 months, mode of travel (nonmotorized or motorized vehicle), whether the participant stayed overnight at work, and frequency of exposure to known potential source of malaria, such as water source and forest.

The presence of clinical symptoms associated with malaria within the past 2 months and the past 24 hours was investigated, including headache, body ache, nausea, vomiting, abdominal discomfort, decreased appetite, fatigue, fever with chill and rigor, and fever. Oral/axillary temperature was taken and recorded at baseline and during follow-up visits, and fever was defined as temperature > 99.5°F. All study procedures were approved by institutional review boards at Duke University Medical Center (Protocol No. Pro00091895), Department of Medical Research, Myanmar, the Defense Services Medical Research Center, Myanmar and the National Institute of Parasitic Diseases, China CDC.

Sample collection.

For the diagnosis of patent P. falciparum malaria infection, finger prick blood was tested by WHO-qualified RDT (Alere Malaria Ag P.f, Standard Diagnostics Bioline Inc., Gyeonggi-do, Republic of Korea) at baseline and any time during study follow-up as clinically indicated. Finger prick blood was spotted onto Whatman 903 and 3MM filter paper at baseline, during scheduled follow-up visits and unscheduled visits for ill symptoms, air-dried, and labeled with a unique participant identification number, initials of blood collector, and date and time of blood collection. These dried blood spot (DBS) samples were stored in individual sealed plastic bags with desiccant in a temperature-controlled facility until transported to central laboratories for molecular analysis.

Study follow-up.

For participants who were RDT-positive at baseline, venous blood (up to 5 mL) was collected and stored in liquid nitrogen for future genotyping of malaria parasites. Referral was made, appropriate treatment by a malaria care provider following the national treatment guideline was ensured, and the treated participant was discharged from the study.

Participants who were RDT-negative and PCR-positive at visit #1 were followed for up to five additional visits (visit #2 to visit #6). In Laiza and Ann, the interval between visit #1 and visit #2 was 4 weeks, followed by biweekly interval for visits #3 to visit #6. In Monywa, the interval between visit #1 and visit #2 was 2 weeks, followed by 5 weekly follow-up visits. Participants who were RDT-negative and PCR-negative at baseline were tested again at visit #2, one week later in Monywa, and 2 weeks later in Laiza and Ann. Therefore, the total maximum number of observations for PCR-negative participants was two and for PCR-positive participants was six. The difference in the follow-up interval was due to logistical constraints for the military population at Monywa where lengthy follow-up (longer than 6 weeks) was not feasible.

Laboratory methods.

From Whatman 3MM DBS samples, RNA and DNA were extracted using a β-mercaptoethanol-spiked guanidine thiocyanate-isopropanol buffer and purified using glass filter plates as previously described.¹⁵ Real-time reverse transcriptase usPCR analysis was performed following published methods targeting the 18S rRNA gene and transcripts for both P. falciparum and P. vivax, with an estimated LoD of 16 parasites/mL for P. falciparum.^5,15,16 A β-tubulin internal control was amplified and detected for all samples to confirm successful DBS extraction. Samples were positive if duplicate measurements for a target had Cq values below 37. DBS samples collected in the Ann and Monywa study sites were analyzed in molecular laboratories located in Yangon and Naypyitaw, Myanmar, respectively, and those from Laiza were analyzed in a malaria laboratory in Shanghai, China, with technical support and coordination provided by a reference laboratory in the United States. All laboratories used identical Roche LightCycler 96 instruments and followed a defined standard operating procedure (SOP). Results were exported and processed using Microsoft Excel software before being merged into a final dataset together with the demographic and survey data. Consistency of methods across the participating laboratories and strict adherence to the SOPs were assured by prestudy training and ongoing refresher training provided by the reference laboratory. The usPCR results were reviewed in real time by experienced molecular analysts who ensured that all laboratory and process control criteria defined in the SOP were met. Subsets (10–20%) of samples from each country laboratory were selected randomly and shipped to the reference laboratory for independent quality control by a team blinded to initial results. Quality-assured usPCR results were shared in real time with the local study teams to guide follow-up activities.

Statistical methods.

Statistics describing sociodemographic characteristics, malaria risk factors, and the history of clinical symptoms were computed and stratified by study site to characterize differences by geography. The sample baseline prevalence (henceforth referred to as “prevalence”) of P. falciparum and P. vivax were calculated by study site and subsite (by village or community), with exact binomial 95% CI, to visualize the heterogeneity in prevalence. Heterogeneity was quantified by the calculation of intraclass correlation coefficient (ICC) on village-level proportions using the analysis of variance (ANOVA) method.¹⁷ Baseline prevalence was also calculated by levels of categorical risk factors to aid interpretation of regression estimates.

Associations between sociodemographic characteristics, known malaria risk factors, and clinical symptom history and malaria positivity (P. falciparum, P. vivax, and all malaria) at screening were ascertained using individual modified Poisson regressions¹⁸ for each candidate variable. Exponentiated regression coefficients were interpreted as prevalence ratios (PRs). To account for sampling design, each regression was adjusted with indicator variables for study site. Each regression was also adjusted for potential confounding by age using dummy indicator variables representing age categories 0.5–16 years, 17–40 years, and 41+ years. Because the regression analyses were considered to be exploratory, no further adjustment for potential confounding was performed to reduce the risk of collider stratification bias and to interpret the regression parameters as age-adjusted total effects.

A similar approach was used to estimate factors associated with counts of positive malaria tests for those who had positive tests at baseline and during the follow-up period. Counts of positive tests did not distinguish between P. falciparum or P. vivax, but rather used baseline diagnosis of P. falciparum/P. vivax as a covariate to ascertain whether the species detected at baseline was associated with total counts of positive usPCR tests for malaria. To account for missing data during follow-up (e.g., missing samples or individuals who were either lost to follow-up after screening, had intermittent missing data, or who dropped out of the study after subsequent follow-up visits), a multiple imputation approach was used to fill in missing values that occurred at any point during follow-up and then to create a summary count of positives based on varying numbers of imputed values (i.e., super-varying variables). Covariates used to impute missing values for malaria diagnosis were data collection site, baseline indicators for P. falciparum and P. vivax (not mutually exclusive, owing to the occurrence of mixed infections with both species), age, occupation, hemoglobin result, distance to work/school, season occupation, occupation type (indoor versus outdoor), clinical symptoms (fever, chill, rigor, nausea, abdominal discomfort, headache, and body ache), female gender, regular use of mosquito net, measured body temperature, and season of data collection (rainy versus dry season).

Ten imputation datasets were produced for regression analysis, which used Poisson regression with robust standard errors and Rubin’s rules for combining results obtained from multiple imputation.¹⁹ Regression coefficients were exponentiated to give the interpretation of incident rate ratios (IRRs). Statistical inference focused on the magnitude and precision of PR and IRR estimates. Because of differences in sample populations and follow-up timing for Monywa, a sensitivity analysis was performed using data only from Laiza and Ann to assess consistency of findings for both baseline subclinical malaria positivity and counts of malaria positive tests over the follow-up time period. In the analysis of variables correlated with the count of positive tests over the 12-week period (or 6-week period for Monywa), indicator variables were added for species of P. falciparum and P. vivax (not mutually exclusive). Because this study was exploratory, no formal power calculation or statistical significance testing was performed.

Baseline prevalence of malaria.

Table 2 reports the proportion of malaria at baseline by RDT (P. falciparum only) and usPCR (P. falciparum and P. vivax) by site. Of 2,705 enrolled, only 5 were positive for P. falciparum by RDT, all at the Ann site in Rakhine State. All RDT-positive participants also tested positive by usPCR. Using usPCR as a gold standard, the estimated specificity of the RDT for detecting P. falciparum was 100% (95% CI: 99.9–100.0%) with a sensitivity of the RDT of 2.4% (95% CI: 0.8–5.5%). By usPCR, malaria prevalence at baseline overall was similar (9.6%) in Laiza and Ann. All infections detected in Laiza were P. vivax, while P. falciparum prevalence was higher than P. vivax in Ann. Monywa had the lowest overall prevalence, with predominantly P. vivax infections. A small number (N = 7) of mixed P. falciparum and P. vivax infections was present only at the Ann site. Results of laboratory quality control procedures showed strong agreement between samples tested at site locations and Duke University. Specifically, at Department of Medical Research (DMR) (N = 96 samples), simple concordance was 96% (Cohen’s Kappa = 0.933) and at Defence Services Medical Research Centre (DSMRC) (N = 63), simple concordance was 95% (Cohen’s Kappa = 0.846).

Within each of the three study sites, usPCR-positive malaria prevalence varied among villages or communities. The range of prevalence in Laiza was 4.6% (95% CI 2.3–8.1) to 13.8% (95% CI 10.8–17.3) (Figure 3, Supplemental Table 2). In Ann, the range was 7.5% (95% CI 2.8–15.6) to 13.0% (95% CI 8.7–18.5). In Monywa, prevalence ranged from 1.3% (95% CI 0.03–6.7) to 4.0% (95% CI 1.5–8.6). To assess heterogeneity in village-level prevalence of P. falciparum, P. vivax, and mixed P. falciparum/P. vivax infections, ICC estimates were compared between the three different infection states. The highest between-village heterogeneity was observed for P. vivax (ICC = 0.028, 95% CI 0.001, 0.055) and the lowest was observed for all malaria (ICC = 0.023, 95% CI: 0.000, 0.046).

View Full Size

Plasmodium falciparum and P. vivax baseline sample prevalence with 95% CIs, by study site subarea. This figure appears in color at www.ajtmh.org.

Citation: The American Journal of Tropical Medicine and Hygiene 107, 3; 10.4269/ajtmh.22-0027

• Download Figure
• Download figure as PowerPoint slide

Temporal dynamics of malaria infection.

Of the 2,705 enrolled, 2,497 participants (92.3%) tested negative by usPCR at baseline (visit #1), with 2,249 (90.1%) remaining PCR-negative at follow-up visit #2. Fifty-three (2.1%) who tested negative at visit #1 were positive by usPCR at visit #2, and 195 (7.8%) who were negative at visit #1 were lost to follow-up at visit #2.

Of the 208 participants who were positive by usPCR at baseline, data for all six study visits were available for 146 participants (70.2%). In Laiza and Ann, 72.4% of P. falciparum or P. vivax positive at visit #1 were still positive at visit #2 one month later, and 37.1% of P. falciparum or P. vivax positive participants remained positive at all six study visits (Figure 4). In Monywa, 46% of PCR-positive at visit #1 remained positive at visit #2, 2 weeks later, whereas 33.3% were positive at all six weekly follow-up visits.

View Full Size

Trajectories of diagnosis by visit (alluvial), by study site. This figure appears in color at www.ajtmh.org.

Citation: The American Journal of Tropical Medicine and Hygiene 107, 3; 10.4269/ajtmh.22-0027

• Download Figure
• Download figure as PowerPoint slide

For the 146 participants who were positive by usPCR at baseline, the patterns of positive and negative tests were relatively equally split between participants who remained PCR-positive for all study visits (N = 54, 37.0%), participants who tested negative at least once and subsequently remained negative (N = 45, 30.8%), and those who were intermittently positive and negative (N = 47, 32.2%) (Supplemental Table 3). Persistence of infection at all six study visits was found in a higher number of participants at the Ann site (N = 33, 41.8%) than the other sites, and those whose initial PCR results were positive for P. falciparum tended to remain positive throughout the study (N = 17, 48.6%) compared with those who tested positive for P. vivax (N = 40, 35.1%). No differences were seen in the proportion of participants at the different sites who were intermittently positive, comparing baseline P. falciparum and P. vivax infections.

Risk factors of subclinical infection.

At baseline, malaria infections were most common among refugees (15.8%), plantation workers (14.6%), and farmers (11.2%) and tended be more common among those with outdoor occupations (10.6% versus 5.8% for indoor occupations) (Supplemental Table 4). Those with symptoms suggestive of malaria both at enrollment and over the previous 2 months had slightly higher rates of baseline malaria infection compared with those who did not have symptoms. Likewise, those reporting more frequent trips to forest and water saw higher baseline prevalence of malaria compared with those reporting frequency of chores as rare or never. At baseline, 3.9% of “dependents” had subclinical infection, among the lowest malaria prevalence of all occupation categories. Plantation workers had the highest relative difference in prevalence of subclinical infection for any occupation category compared with dependents (PR = 4.09, 95% CI 2.23, 7.51), after adjustment for site and age (Table 3). Among all nonreferent occupation categories, students had moderately higher rates of positive malaria tests over the follow-up time period, compared with dependents (IRR = 1.46, 95% CI 1.08, 1.97). Those with occupations classified as occurring predominantly outdoors were more likely to test positive for malaria at baseline, with unadjusted baseline prevalence for outdoor occupation of 10.6% versus 5.8% for those with indoor occupations (PR = 2.57, 95% CI 1.84, 3.58); however, those with outdoor occupations also saw fewer subsequent positive tests (IRR = 0.66, 95% CI 0.55, 0.80) than individuals with indoor occupations who tested positive at baseline.

Older age groups were more likely than children aged 6 months to 16 years to have subclinical malaria at baseline (PR = 1.59, 95% CI 1.13, 2.25), although the older age groups had fewer positive counts over the follow-up time points than younger children who had tested positive at baseline (IRR = 0.80, 95% CI 0.65, 0.99 for individuals aged 17–40 years, IRR = 0.74, 95% CI: 0.59, 0.91 for those aged 41 years or older).

Those who traveled farther in a typical day were more likely to have subclinical malaria at baseline, with rates of positivity increasing with increasing distance traveled (PR = 1.83, 95% CI 1.35, 2.50 for those traveling 1–5 km; PR = 2.60, 95% CI 1.67, 4.04 for those traveling > 5 km) although travel history did not correlate with the subsequent number of positive tests.

Individuals who performed chores involving frequent trips to water sources and forests were more likely to test positive for malaria at baseline (PR = 2.00, 95% CI 1.39, 2.86 for trips to water; PR = 1.85, 95% CI 1.39, 2.43 for trips to forest). Subsequent counts of positive tests were modestly correlated with lower frequency of trips to forest (IRR = 1.23, 95% CI 1.05, 1.44) but not strongly correlated with lower frequencies of trips to water.

The time of year during which screening occurred did not correlate well with either the probability of being positive at baseline (PR = 1.09, 95% CI 0.83, 1.43 for screening during non–peak season versus peak season reference) or the subsequent number of positive tests for those who tested positive at baseline (IRR = 0.92, 95% CI 0.79, 1.08).

Individuals who tested positive for P. falciparum at baseline tended to have higher subsequent counts of positives than those who tested positive only for P. vivax (IRR = 1.34, 95% CI 1.11, 1.63). Self-reported symptoms suggestive of malaria in the 2 months preceding screening were not correlated with the probability of testing positive at baseline, although reports of body ache/pain (IRR = 1.29, 95% CI 1.14, 1.46) and vomiting (IRR = 1.44, 95% CI 1.04, 1.99) were associated with higher counts of positives tests for those testing positive at baseline.

Results remained largely unchanged in a sensitivity analysis when data from Monywa were excluded (Supplemental Table 5), with only modest changes in estimated magnitudes of association between measured temperature and subclinical malaria at baseline and the counts of subsequent positive malaria results for individuals reporting abdominal discomfort in the 2 months preceding baseline screening.

Overall, those potential risk factors that were associated with testing positive for subclinical malaria at baseline were not associated with the count of positives that occurred after baseline. In several cases, factors predicting initial infection produced effects on counts of positives that were in the opposite direction (e.g., age, trips to forest, and indoor/outdoor occupation type).