The 2017 School Malaria Parasitological Survey (SMPS) was a cross-sectional study of children who were 5–16 years of age and enrolled in public primary schools in mainland Tanzania. Methods for site selection and survey design mirrored the 2015 SMPS and have been previously described (22). Study regions were selected through a multistage sampling scheme to maintain geographic representation and reflect the heterogeneity of malaria transmission across Tanzania (22,26). The number of schools randomly selected per region was proportional to each region’s respective population (22,26). Within each school, an average of 100 students were randomly selected for screening. After consent, each student was interviewed to obtain demographic and clinical characteristics, a malaria RDT was performed, and a dried blood spot (DBS) sample was collected (22,26). The survey largely coincided with each region’s rainy season. From among students who provided a DBS, we selected a stratified random subpopulation for nonfalciparum malaria testing. To maintain representativeness, we selected students in proportions that equaled regional proportions reflected within the broader survey population.

Informed consent had been obtained from students and their legal guardians before survey data or blood sample collection, and ethical clearance was given by the Tanzania National Institute for Medical Research. Analysis of de-identified samples was approved by the Institutional Review Board of the University of North Carolina, Chapel Hill (approval no. 19-1495).

During the survey, malaria detection was conducted by using CareStart Malaria Pf/PAN (HRP2/pLDH) Ag Combo RDTs (AccessBio, https://www.accessbio.net) that were specific for P. falciparum HRP2 and pan-pLDH antigens. RDTs were considered positive if they were positive for either antigen. Schools and councils were grouped into epidemiologic malaria transmission risk strata on the basis of P. falciparum prevalences in children estimated from the 2014–15 Tanzania SMPS (22,26). P. falciparum prevalence was defined as very low if <5%, low if 5 to <10%, moderate if 10 to <50%, and high if >50% (22,26). DBS samples collected on Whatman filter paper (Cytiva, https://www.cytivalifesciences.com) were shipped to the University of North Carolina (Chapel Hill, NC, USA) for molecular testing.

We extracted DNA from three 6-mm punches from each DBS sample by using a Chelex method (27) and performed real-time PCR targeting the 18S rRNA subunit of malaria as previously described (28) (Appendix Table 1). We performed PCR for each Plasmodium spp. independently with appropriate controls. We prepared positive controls for P. falciparum detection by using whole human blood and cultured P. falciparum strain 3D7 parasites (BEI Resources, https://www.beiresources.org) to create mock DBS samples and for nonfalciparum species detection by using plasmid DNA (BEI Resources). We serially diluted the control samples and extracted DNA as described. We estimated semiquantitative parasitemias for nonfalciparum species by assuming 6 18S rRNA gene copies/parasite (28) and multiplying by 4.0 to account for the 4-fold dilution of blood: ≈26 µL blood from 3 DBS punches (29) in 100 µL final volume of eluted DNA. We performed 40 PCR cycles for P. malariae and P. falciparum and 45 PCR cycles for P. ovale and P. vivax to enable detection of low-density infections (28). We previously validated this approach by using 390 negative controls comprising water (n = 22) and human DNA (n = 368) and >170 positive controls with decreasing nonfalciparum parasite densities; no false-positives were detected (28). We assessed PCR specificity by testing against 10 controls from each of the other Plasmodium spp.; no false positives were detected (Appendix Table 2). Our laboratory at the University of North Carolina participates in the World Health Organization malaria molecular quality assurance scheme, identifying and determining Plasmodium spp. in blinded samples every 6 months, and has consistently achieved high marks for assay performance across species. In this study, we did not detect false-positive amplification among 20 negative controls per each species-specific assay (Appendix Table 3). We performed further real-time PCR on a subset of P. ovale–positive samples to distinguish between P. ovale wallikeri and P. ovale curtisi (30,31). To evaluate potential bias from differences in PCR cycle numbers between species, we conducted a sensitivity analysis of randomly selected students (n = 750) stratified by malaria transmission risk. We performed semiquantitative real-time PCR of the 18S rRNA gene to 45 cycles to detect P. falciparum and P. malariae infections.

We calculated overall malaria species-specific prevalences and prevalence of single- and mixed-species infections. We did not adjust prevalences for sampling weight because nonfalciparum samples were selected randomly and in equal proportion to the broader survey sample.

We performed descriptive statistical analyses of student characteristics according to Plasmodium spp. We analyzed differences between P. falciparum and nonfalciparum single-species infections by using Pearson χ² and Kruskal-Wallis rank-sum tests assuming nonnormality and applied Fisher exact test for small frequency counts. We performed similar analyses to compare malaria-positive and -negative students according to Plasmodium spp. Missing data were summarized, but we performed analyses on nonmissing data only.

We assessed regional variation in prevalence of each species through geospatial mapping by council and region. We aggregated numbers of infections and students by council and estimated and mapped council-level prevalences for each species. We calculated scaled prevalences by dividing the proportion of each council’s prevalence by the highest council prevalence for each Plasmodium species, as follows (Figure 6):

where P is the prevalence for a given council, d_i. We calculated and mapped differences between scaled nonfalciparum and scaled P. falciparum prevalences for each council. This method compared prevalence estimates between each nonfalciparum species and P. falciparum, while accounting for differences in the absolute burden of each species.

We performed analyses by using R version 4.0.2 (The R Project for Statistical Computing, https://www.r-project.org) and used the eulerr (https://cran-r-project.org/package=eulerr) and sf version 0.9–7 (32) packages for prevalence visualization and mapping. We sourced shapefiles from the Global Administrative Areas database (https://gadm.org) and collected elevation measurements from the US National Aeronautics and Space Administration, Shuttle Radar Topography Mission (https://www.nasa.gov).

We selected a total of 3,456 students from 180 schools across 8 geographic regions for nonfalciparum malaria testing from among 17,131 students in the SMPS who had available DBS samples. We did not detect differences in student characteristics between those in the nonfalciparum malaria and SMPS DBS populations (Appendix Table 4). Median (interquartile range [IQR]) student age in the nonfalciparum study population was 11 (9–13) years; distribution of male (51%) and female (49%) students was similar. Malaria dual-antigen RDTs were positive in 20% of students. Most students attended schools in regions classified as high (51%) or moderate (13%) malaria transmission risk (Table 1).

We identified P. falciparum infections in 22% (95% CI 21%–23%, n = 755), P. ovale in 24% (95% CI 22%–25%, n = 814), P. malariae in 4% (95% CI 3%–5%, n = 136), and P. vivax in 0.3% (95% CI 0.2%–0.6%, n = 11) of students, including single- and mixed-species infections (Appendix Table 5). Most (64%, n = 519) P. ovale infections were single-species infections; 28% (n = 224) were co-infections with P. falciparum only (Figure 1). Conversely, most (40%, n = 55) P. malariae infections were co-infections with both P. ovale and P. falciparum; 32% (n = 44) were co-infections with P. falciparum only. We determined 36% (n = 4) of P. vivax infections were single-species infections, and 43% (n = 326) of P. falciparum infections were co-infections with >1 nonfalciparum malaria species.

We conducted a sensitivity analysis, detecting P. falciparum and P. malariae by using PCR cycle thresholds of <45 to evaluate different PCR cycles between assays. We observed 25% (95% CI 21%–29%) P. falciparum and 3% (95% CI 2%–5%) P. malariae prevalences, weighted according to student distribution within the total nonfalciparum population by transmission risk strata (Appendix Table 6). Within that subset, 2.5% (n = 4) of P. falciparum and 10% (n = 2) of P. malariae infections were detected at cycle thresholds of 40–45. Thus, >97% of P. falciparum and 90% of P. malariae infections were detectable by the primary 40-cycle assay in our study.

We evaluated differences in student characteristics according to Plasmodium spp. infection (Table 1; Appendix Table 7). We detected P. ovale single-species infections more frequently than P. falciparum infections in slightly younger (median 11 vs. 12 years of age; p<0.001) and female (54% vs. 45%; p = 0.009) students. Comparing RDT sensitivity to PCR, we observed 8% (n = 40) of students with P. ovale single-species infections were RDT-positive for any band, whereas 33% (n = 8) of those with P. malariae and 69% (n = 295) with P. falciparum single-species infections were RDT-positive. Co-infections with P. falciparum and nonfalciparum were RDT-positive in 78% (n = 253/325) of cases detected by PCR. Although only 3% (n = 13) of P. falciparum single-species infections and no P. malariae or P. vivax single-species infections were detected in low transmission risk strata, 35% (n = 181) of P. ovale single-species infections occurred in regions classified as low or very low malaria transmission risk. High epidemiologic risk strata harbored most single-species infections across all 4 Plasmodium spp. and also mixed infections with P. falciparum.

Malaria parasitemia estimated by semiquantitative PCR was low across nonfalciparum species (Figure 2). Median (IQR; min–max) P. ovale density was 7.2 (1.3–25.0; 0.1–168,596) parasites/µL, comparable to P. malariae density at 11.7 (2.7–54.9; 0.3–1,214) parasites/µL. P. vivax density was ≈0.6 (0.3–0.8; 0.1–8.1) parasites/µL. Although 18% (n = 25) of P. malariae infections had a parasite density >100 parasites/µL, we rarely observed that level for P. ovale (3%, n = 25) and never for P. vivax. P. falciparum density was also low at 13.1 (2.6–55.9; 0.1–8,248) parasites/µL; however, 17% (n = 132) of P. falciparum cases had a parasite density >100 parasites/µL, and 3% (n = 24) had >500 parasites/µL. Median (IQR) density among P. ovale mixed infections was 3.1 (1.2–11.4) parasites/µL and 13.5 (1.3–30.1) parasites/µL for P. ovale single-species infections (p<0.001), whereas densities were similar between single- and mixed-species infections among the other malaria species (Figure 2).

Among 814 samples positive for P. ovale, 60 (7%) samples with the highest parasitemia were selected for PCR to distinguish between P. ovale wallikeri and P. ovale curtisi. Species determination by PCR was successful in 35% (n = 21) of samples; P. ovale curtisi was detected in 17 samples and P. ovale wallikeri in 9 samples. We identified P. ovale curtisi and P. ovale wallikeri co-infections in 5 students. We did not perform further characterization because of limited sample sizes.

We detected P. ovale across all 8 regions sampled in Tanzania, indicating widespread distribution (Table 2; Figures 3, 4). P. ovale prevalence was highest within the northern Kagera (34%, n = 273) and central Tabora (17%, n = 139) regions. We detected P. ovale curtisi infections in 6 of 8 regions (all but Arusha and Rukwa) and P. ovale wallikeri in 5 of 8 (Kagera, Mara, Tabora, Tanga, and Iringa) regions. We observed high prevalence of P. malariae in Kagera (29%, n = 39) and in southernmost Mtwara (28%, n = 38), and P. vivax was predominantly distributed along the northwestern borders of Tanzania in Kagera (55%, n = 6); select, isolated cases of P. vivax were also detected in southern and eastern regions. Arusha and Iringa did not have any cases of P. malariae or P. vivax infections and had the lowest frequencies of P. ovale (3%, n = 23, in Arusha; 4%, n = 30, in Iringa) and P. falciparum (0.4%, n = 3, in Arusha; 0.1%, n = 1, in Iringa) infections.

We detected malaria infections in students who were predominantly located at elevations <1,500 m, including 85% (n = 693) infected by P. ovale, 94% (n = 707) by P. falciparum, 95% (n = 129) by P. malariae, and 100% (n = 11) by P. vivax (Table 2). Most (77%, n = 2,667) students enrolled in our study were from schools located at elevations <1,500 m. Among students located at elevations >1,500, P. ovale infections were detected most frequently in 15% (n = 121) of students compared with 6% (n = 48) infected with P. falciparum, 1% (n = 7) infected with P. malariae, and 0% infected with P. vivax.

We compared scaled prevalence estimates for nonfalciparum species with P. falciparum and identified areas where prevalences were higher than expected for P. ovale and P. malariae on the basis of P. falciparum frequency (Figure 5); P. vivax infections were too infrequent for comparison. In the southern and southwestern highlands and northwestern lake regions (Iringa, Rukwa, Tabora, and Kagera), scaled P. ovale prevalences were higher than P. falciparum prevalences. Scaled prevalence of P. malariae was notably higher than that of P. falciparum in the Karagwe council in Kagera and Mtwara municipal council in Mtwara. In most other areas, scaled prevalence of P. malariae was similar to or lower than P. falciparum prevalence.