Transmission intensity by elevation
Despite the small size of the study area, there were noticeable differences in transmission intensity comparing cohort households above the median elevation (higher elevation) versus below the median (lower elevation). Using data from all children in the cohort through one year, there was lower malaria incidence in children living at higher compared to lower elevation (1.4 versus 1.9 episodes per person-year (ppy) (IRR = 0.73, 95% CI: 0.53–1.00). In addition, the daily human biting rate for anopheline mosquitoes (dHBR) was considerably lower in households at higher versus lower elevations (2.01 vs 8.9, p = 0.08), adding evidence to support higher transmission at lower elevation. Thus, our cohort data suggest modest but meaningful differences in transmission within the study area by both clinical and entomologic measures, with transmission higher in houses at lower elevation.
Study population and genotyping
A total of 408 samples with ≥ 5 parasites/µL of blood from 80 households were genotyped using 26 microsatellite markers. Of these, 349 (85.5%) samples were successfully genotyped at 15 or more loci and included in subsequent analysis; these samples were collected from 230 unique individuals (142 children and 88 adults) (Table 1). Of these, 128 (51.2%) children and 4 (4.0%) adults reported and/or had objective fever at the time of their sampled infections.
Table 1. Descriptive statistics of 349 successfully genotyped P. falciparum infections from the Kihihi Subcounty of Uganda.
| Children (0.5–11 years) | Adults (> 18 years) | Total |
Number of individuals | 142 | 88 | 230 |
Samples (coverage at > = 15 markers) | 250 | 99 | 349 |
Symptomatic infections, n (%) | 128 (51.2%) | 4 (4.0%) | 132 (37.8%) |
Microscopy positive, n (%) | 161 (64.4%) | 5 (5.1%) | 166 (47.6%) |
Monoclonal infections, n (%) | 75 (30.0%) | 9 (7.1%) | 82 (23.5%) |
| Higher Elevation | Lower Elevation | Total |
Unique Households | 50 | 30 | 80 |
Avg number of children/household (mean) | 1.6 | 2.1 | 1.8 |
Average child age in years (mean) | 5.8 | 5.5 | 5.6 |
Monoclonal infections, n (%) | 43 (24.9%) | 39 (22.2%) | 82 (23.5%) |
Mean Heterozygosity | 0.72 | 0.72 | 0.73 |
Mean Multiplicity of Infection | 2.2 | 2.6 | 2.4 |
Complexity and genetic diversity of infections
The majority of genotyped infections were polyclonal (76.5%) with a mean MOI of 2.4 (range = 1–6, Supplemental Fig. 1). Children tended to have less complex infections, with 30.0% monoclonal infections in children compared to only 7.1% monoclonal infections in adults (p = 0.0002, Table 1). Samples from lower elevation households had a higher mean MOI than samples from the higher elevation households (2.6 vs 2.2, p = 0.008); this was driven by differences in MOI between children at lower vs higher elevation (2.4 vs 2.1, p = 0.024). However, populations at both elevations had a similar proportion of monoclonal infections: 24.9% for higher elevation and 22.2% for lower elevation samples (p = 0.56). The overall population level genetic diversity was high in Kihihi, with a mean heterozygosity (He) of 0.73 [range: 0.37–0.91] with no difference between elevations or between adults and children.
Spatial scale of genetic relatedness
Pairwise genetic relatedness was determined between all genotyped samples obtained from different individuals, including those with polyclonal infections, resulting in 60,568 pairwise comparisons. Of these, there were 733 intra-household pairs and 3,306 pairs of monoclonal infections. Infections from the same household were more likely to be genetically related than infections from participants that did not (Fig. 2A); a similar result was seen when limiting this analysis to only monoclonal samples. There was suggestion of a small increase in relatedness between infections from participants in different households living < 2 km vs. >= 2 km apart, though statistical power to evaluate this question was limited due to small numbers. A total of 93 (0.14%) infection pairs were highly related (IBS > = 0.6). Highly related pairs showed significant spatiotemporal clustering compared to those with IBS < 0.6. (Fig. 2B and 2C). 79/93 (84.9%) of highly related infections were from participants who lived within 10 km of each other, and 78/93 (83.9%) of highly related infections occurred within 90 days of each other. These results that highly related infections can be used to explore P. falciparum transmission patterns in this moderate transmission setting.
Within-population and within-household transmission
On a finer spatial scale, we found a higher proportion of highly related infections in members of the same household compared to individuals living in different households (2.9% vs 0.1%, p < 0.0001), suggesting that transmission-related infections cluster within households. To test the hypothesis that highly related infections were more likely to be observed within the same household where transmission was lower, we compared the proportion of highly related infections within households by elevation. As hypothesized, we found a higher proportion of highly related pairs of infections within households at higher elevation compared to lower elevation (6.3% vs 0.9%, p = 0.0005) (Fig. 3A). This result suggests that within-household clustering may occur more commonly and/or be more easily detectable in areas of lower compared to higher transmission.
There were too few within-household highly related infection pairs (n = 21) to evaluate within-household transmission events. We did examine the proportion of highly related infections in child-child, child-adult, and adult-adult paired samples both within and between households. In child-child pairwise comparisons, the proportion of highly related infections was higher than in comparisons for child-adult or adult-adult pairs (Fig. 3), although the greater MOI observed in adults complicated interpretation of this result. However, when we restricted the analysis to monoclonal samples, there was again a higher proportion of highly related samples in child-child pairs than in child-adult pairs (0.5% vs 0.06%, p = 0.005), and no adult-adult pairs were highly related. As might be expected given other evidence for within-household clustering, this pattern was more pronounced within households, with 5.4% of child-child pairs within households highly related compared to 0.2% of child-child pairs from different households (p < 0.0001) (Fig. 3). When the analysis was restricted to monoclonal samples, this pattern held, with 6.8% of child-child pairs within-households highly related compared to 0.4% of child-child pairs from different households (p < 0.0001).