Estimating Annual Fluctuations in Malaria Transmission Intensity and in the Use of Malaria Control Interventions in Five Sub-Saharan African Countries

Ethics.

The study was approved by national independent Ethics Committees and local institutional review boards in Burkina Faso (BF), Ghana (GH), Kenya (KE), Senegal (SN), and Tanzania (TZ), and conducted in accordance with the provisions of the International Conference on Harmonisation and Good Clinical Practice guidelines.

Study population.

Individuals aged 6 months to < 10 years, whose parents or legally acceptable representative had provided informed consent for study participation, were randomly selected each year in each of the study sites (see Study design section) using population listings generated from local Health and Demographic Surveillance Systems (HDSS) and following a stratification by age-group (see the Supplemental Appendix Section 1). Children in care, or actively participating in any trial involving the administration of an investigational malaria vaccine and/or drug, were excluded.

Study design.

Malaria transmission intensity levels are consensus indicators developed by the Global Malaria Eradication Programme to measure malaria endemicity⁷ using a standardized methodology. There are several methods for estimating MTI, including entomological inoculation rates (EIRs), serological conversion rates (SCRs), and blood parasite prevalence. Although EIR is a standard method, the measure is challenging, and the interpretation and comparability of the setting may be difficult because of vector heterogeneity.⁸ The methodology and interpretation of SCRs to classify the intensity of malaria is still not commonly used. Plasmodium falciparum prevalence, despite requiring trained staff for slide reading, provides a standardized and relatively easy to implement method to assess MTI in study sites of varied transmission intensity, and has often been used in previous epidemiological studies.^7,9,10 Therefore, PfPR was the selected method to assess potential variations in MTI in the present study.

The study is multicentric with study sites corresponding to geographically limited catchment areas located in low, moderate, or high MTI regions of SSA, and having an HDSS in place.

Up to nine annual cross-sectional surveys will be conducted during the malaria peak transmission period in each study site (from mid-September to mid-December in Western African sites and from mid-April to mid-August in Eastern African sites). To optimize the detection of the peak transmission, each study site has been equipped with a weather station to record meteorological data such as rainfall, temperature, and humidity. Surveys were conducted during the course of the rainy season preferably when rains decrease, which should correspond to the period of highest malaria transmission. In this article, results of the first two annual surveys are presented. Survey 1 (S1) and survey 2 (S2) were conducted in seven study sites in five SSA countries: BF (Nouna, Saponé), GH (Kintampo), KE (Kombewa), SN (Keur Socé, Niakhar), and TZ (Korogwe). Sites in GH (Kintampo) and KE (Kombewa) are part of the areas where the RTS,S/AS01_E vaccination will be implemented through the national Expanded Programs on Immunization in the framework of the MVIP. It is important to note that this study is also conducted in SSA areas where the RTS,S/AS01_E vaccine will not be introduced in the framework of the MVIP because GSK initiated the study before the WHO recommendation.^5,11

Data collection.

Demographic details (age and gender), medical history, and information on care-seeking behaviors (hospitalization for malaria within the last 3 months, visits to healthcare provider for fever or malaria treatment in the previous 14 days, anti-malaria therapy received within the last 14 days); malaria control intervention usage (bednets [new {not older than a year}, pierced/torn, impregnated], indoor residual spraying [IRS]); the usage of coils, repellents, and local herbs; and malaria potential risk factors (rural/urban area, house construction materials, use of electricity, and open/closed water source) were recorded for all participants at the time of the survey. Axillary body temperature was measured and recorded during the survey visit.

To assess within-site heterogeneity of PfPR, study areas were mapped by villages using grid referencing and divided into 3–14 segments with a minimum of 10 enrolled individuals per segment. Segments will remain unchanged for the duration of the study.

Assessment of parasitemia.

Both microscopy and Nucleic Acid Amplification Tests (NAATs) were used in parallel on all participants to assess parasitemia. The latter are expected to be more sensitive and specific, particularly in cases of low parasite density.^12,13 In brief, a blood sample was collected by finger prick for thin and thick blood films for the microscopy assessment and filter paper blood spots for NAATs. Blood smears were examined by two local independent microscopists, and any discrepancies were settled by a third reader. Parasitemia was measured by the examination of 100 high-powered fields on thick smear to determine the presence of parasites; in the case of a positive result, additional 100 fields were examined to assess the presence of multiple species. Plasmodium species and sexual forms were identified on thin blood film. Parasite density was computed as the geometric mean of two readings, counting parasites against 200 white blood cells on thin blood film, assuming 8,000 white blood cells/μL. Parasite density was categorized as low (< 2,500 parasites/µL), medium (2,500–9,999), high (10,000–19,999), or very high (> 20,000). In the case of low density (< 10 parasites against 200 leukocytes), parasite count was conducted against 500 white blood cells. The parasite count technique was replicated to count gametocytes. In parallel with microscopy, asexual and sexual parasitemia was assessed by NAATs using both real-time quantitative PCR (QT-PCR) assay and real-time nucleic acid sequence-based amplification (QT-NASBA) assay. Quantitative PCR allowed detection of asexual and sexual parasites combined, the final output being qualitative (positive or negative) and semi-quantitative (high, medium, low, and negative). QT-NASBA allowed detection of gametocytes, the final output being qualitative (positive or negative). Details for blood slide and NAAT assessment of parasitemia are available in the Supplemental Appendix (Sections 2 and 3, respectively).

If fever (i.e., axillary temperature ≥ 37.5°C) was recorded at the time of the visit or reported to have occurred within 24 hours before the visit, a malaria rapid diagnostic test (RDT) was conducted using blood from the finger prick sampling. If the RDT was positive, then treatment was given according to national guidelines. Moreover, any participant identified as being parasite positive following microscopy was traced to receive treatment according to national guidelines.

Statistical methods.

The planned sample size was 600 participants per study site and per survey distributed as 400 participants between the ages of 6 months to 4 years and 200 participants between the ages of 5 and 9 years. The sample size was calculated to ensure sufficiently narrow CIs around study site–specific PfPR estimates (with a maximum residual standard error of 0.25).

Plasmodium falciparum parasite prevalence and prevalence of gametocytes were estimated as the proportion of participants infected, or carrying gametocytes, respectively, among participants tested. Prevalences were estimated by age and by site. The agreement between the two diagnostic methods used in the framework of this study (parasitemia measured by microscopy versus NAATs) was described using the Cohen’s kappa coefficient and assessed using the Landis and Koch scale.¹⁴ The prevalence of Plasmodium species other than P. falciparum was estimated as the proportion of infected participants among participants tested. The within-site heterogeneity between segments was tested using Cochran’s Q-test based on inverse variance weights.

Malaria control intervention coverage was estimated as the proportion of users among participants for which this information was available. The care-seeking behaviors (treatment sought for malaria or fever in the 14 days before the visit and hospitalization for malaria in the last 3 months before the visit) were described as the proportion of participants having sought for health care among all participants. In addition, a risk factor analysis for malaria infection (dependent variable: P. falciparum parasitemia as measured by microscopy) was conducted using a multivariable logistic regression with study site as cluster and using a backward strategy for the selection of significant explanatory variables, that is, predefined potential risk factors and/or the use of malaria control interventions (Supplemental Appendix Section 4). Age was computed as a continuous variable.

Year-to-year variation in P. falciparum prevalence.

Plasmodium falciparum parasite prevalence estimated using microscopy varied between study sites in both surveys, with the lowest prevalence figures in Senegalese sites and the highest in BF (Table 1, Figure 2). In Kintampo and Kombewa sites that are located in the MVIP areas, PfPR in children aged 6 months to 4 years ranged from 24.8% to 27.3% depending on the study site and the survey. In both surveys and across all sites except in SN, PfPR was lower in the 6-month to < 5-year than in the 5-year to < 10-year age-group. Plasmodium falciparum parasite prevalence was similar in S1 and S2, except for a higher prevalence in S2 in participants from the 6-month to < 5-year age-group in BF Nouna and lower prevalence in S2 in both the 6-month to < 5-year and 5-year to < 10-year age-groups in participants from TZ Korogwe. Significant within-site heterogeneity was detected in all sites, except in SN: BF Nouna (S1 P < 0.0001, S2 P = 0.0184), BF Saponé (S1 P = 0.0321, S2 P < 0.0001), TZ Korogwe (S1 P = 0.0037, S2 P = 0.0275), KE Kombewa (S1 P = 0.0220, S2 P < 0.0001), and GH Kintampo (S2 only P = 0.0065).

View Full Size

Plasmodium falciparum parasitemia prevalence measured by microscopy, by age category, and by site. 6 M to < 5 Y = 6 months to younger than 5 years; 5 Y to < 10 Y = 5 years to younger than 10 years. Error bars depict 95% CI.

Citation: The American Journal of Tropical Medicine and Hygiene 103, 5; 10.4269/ajtmh.19-0795

• Download Figure
• Download figure as PowerPoint slide

Similar trends were observed when measured by QT-PCR, with PfPR varying from 2.7% in SN Niakhar to 69.9% in BF Nouna (not measured in SN Keur Socé) in S1 and from 1.3% in SN Keur Socé to 76.3% in BF Nouna in S2 (Supplemental Table 3).

The prevalence of gametocytes measured by microscopy ranged from 0.5% to 4.0% in S1 and from 0.0% to 5.3% in S2 in all sites, except in BF (15.8% and 22.7% in S1 and 12.8% and 18.2% in S2 in Nouna and Sapone, respectively). Among participants tested positive for asexual parasites, 7.5–66.7% carried gametocytes in S1 (Kintampo and Keur Socé, respectively) and 0.0–83.3% in S2 (Niakhar and Keur Socé, respectively) (Table 2). The proportion of infected participants carrying gametocytes as estimated by QT-NASBA ranged from 13.3% in SN Niakhar to 59.2% in BF Sapone in S1 (not measured in Keur Socé) and from 7.1% in SN Niakhar to 66.7% in SN Keur Socé in S2 (Supplemental Table 3).

Agreement between diagnostic tests.

Across surveys, approximately 21.6% of participants with a positive QT-PCR result had a negative microscopy reading (285 of 1,268 positive participants per QT-PCR in S1 and 296 of 1,417 in S2) and around 8.8% of participants positive for microscopy had a negative result with QT-PCR (92 of 1,075 positive participants per microscopy in S1 and 110 of 1,231 in S2) (Table 3). The proportion of participants with a negative result by microscopy among participants with a positive QT-PCR ranged between 42.9% and 78.6% in sites with low MTI (Niakhar, Keur Socé, and Korogwe).

Across all sites, Cohen’s kappa coefficient between qualitative results of the two methods of measurement of parasitemia (microscopy versus QT-PCR) using the Landis and Koch scale showed a substantial agreement in both surveys (S1: kappa = 0.75 [95% CI: 0.73; 0.78]; S2: kappa = 0.78 [95% CI: 0.75; 0.80]). Cohen’s kappa coefficient between semi-quantitative results for parasitemia measured by microscopy versus QT-PCR also showed a substantial agreement in both S1 (kappa = 0.65 [95% CI: 0.63; 0.68]) and S2 (kappa = 0.66 [95% CI: 0.64; 0.68]).

Between 25.0% (192 of 767 in S2) and 33.8% (185 of 548 in S1) of participants tested positive for gametocytes by QT-NASBA were also detected positive by microscopy (Supplemental Table 4).

Prevalence of Plasmodium species other than P. falciparum.

Infection with Plasmodium malariae was observed in 1.5% of participants in S1 and 3.0% of participants in S2 (Supplemental Table 5). Coinfection with both P. malariae and P. falciparum was more frequent than single infection with P. malariae alone in both surveys (S1: 4.6% versus 0.2%; S2: 7.4% versus 1.2%). Of the 62 participants in S1 and 128 participants in S2 infected with P. malariae, 55 (88.7%) and 91 (71.1%) were also infected with P. falciparum, respectively. Infection with Plasmodium ovale was low in both S1 and S2 (0.5% and 0.2% of participants). Plasmodium vivax was not observed in S1 and in only one participant in S2. Infections with species other than P. falciparum were mostly observed in sites of medium-to-high P. falciparum prevalence.

Year-to-year variation in the use of malaria control interventions.

Overall, 89.5% and 86.4% of children used a bednet the night before the survey in S1 and S2, respectively (Table 4, Supplemental Table 6). The highest use of bednets was in KE Kombewa (97.5%) in S1 and TZ Korogwe (99.2%) in S2, and the lowest in GH Kintampo (70.2%) in S1 and SN Niakhar (69.6%) in S2 (Table 4). A decrease in usage between the two surveys was observed in Kombewa, Keur Socé, and Niakhar, whereas an increase was observed in Kintampo and Korogwe.

Overall, 70–80% of the bednets were impregnated, 60–70% were new, and approximately 25% were torn (Supplemental Table 6). Details for the characterization of bednet usage (new, impregnated, and pierced/torn) by study site are shown in Figure 3.

View Full Size

Bednet characterization according to obsolescence, impregnation, and condition, by survey and by site. impregnated bednet = bednet dipped in an insecticide liquid before or after purchase; new bednet = bednet not older than 1 year; S1 = survey 1; S2 = survey 2. Error bars depict 95% CI.

Citation: The American Journal of Tropical Medicine and Hygiene 103, 5; 10.4269/ajtmh.19-0795

• Download Figure
• Download figure as PowerPoint slide

Participant’s recall of IRS in the past 12 months was recorded for a very low number of participants (across surveys, 4.1% overall), mainly in SN (Supplemental Table 7).

Overall, usage of coils and repellents was limited, around 10% of the population in both surveys with variations per site between 1% and 40% according to the survey (Supplemental Table 7).

Year-to-year variation in reported fever and care-seeking behaviors.

Fever in the last 24 hours was reported for approximately a quarter of the participants in both surveys, with differences across study sites ranging from 3.6% in BF Saponé to 63.8% in KE Kombewa in S1 and from 3.0% in SN Keur Socé to 71.6% in KE Kombewa in S2 (Supplemental Table 8). Across all sites, occurrence of fever was higher in P. falciparum–infected versus non-infected participants (35.6% versus 21.1% in S1 and 33.5% versus 22.5% in S2, respectively). In both surveys, fever was more frequently reported by participants with higher parasite densities (Supplemental Table 9).

In S1, 15.7% of participants had sought treatment against malaria or fever in the 14 days before the survey compared with 12.8% in S2, ranging from 0.0% to up to 33.3% depending on the study site and on the survey (Table 5, Supplemental Table 10). Plasmodium falciparum–infected children sought treatment against malaria or fever more often than non-infected children (20.1% versus 14.0% in S1 and 20.0% versus 9.8% in S2, respectively).

The proportion of participants hospitalized for malaria was 2.6% in S1 and 2.8% in S2, with no marked difference between P. falciparum–infected and non-infected participants 2.8% versus 2.5% in S1 and 3.4% versus 2.6% in S2, respectively; (Table 5, Supplemental Table 10).

Association between potential risk factors and P. falciparum infection.

An exploratory multivariable model was used to assess the association between potential risk factors or malaria control interventions and P. falciparum infection. Across both surveys, houses equipped with electricity (odds ratio [OR]: S1 0.75 [95% CI: 0.61; 0.93]; S2 0.89 [95% CI: 0.80; 0.98]), cement/plaster walls versus mud (OR: S1 0.81 [95% CI: 0.69; 0.96]; S2 0.87 [95% CI: 0.76; 0.99]), and nets on all windows (OR: S1 0.73 [95% CI: 0.63; 0.84]; S2 0.79 [95% CI: 0.66; 0.94]) were associated with a lower risk of infection with P. falciparum. In addition, older age was associated with a higher risk of infection (OR: 1.14 [95% CI: 1.05; 1.24] in S1 and 1.09 [95% CI: 1.02; 1.16] in S2; Supplemental Tables 11 and 12).

Figure 4 represents a plain language summary, which elaborates on the epidemiologic study relevance that could be shared with patients by healthcare professionals.

View Full Size

Plain language summary.

Citation: The American Journal of Tropical Medicine and Hygiene 103, 5; 10.4269/ajtmh.19-0795

• Download Figure
• Download figure as PowerPoint slide

Characterizing MTI in different SSA settings.

Using PfPR as a proxy, this study aims at characterizing MTI¹⁵ in different SSA settings, including areas in GH and KE where the RTS,S/AS01_E malaria vaccine is currently introduced in the framework of the MVIP. More specifically, considering the WHO recommendation to operate the MVIP in moderate to high transmission areas of SSA, PfPR in children aged 6 months to 4 years in Kintampo and Kombewa sites was high, ranging from 24.8% to 27.3% depending on the study site and the survey. Similar PfPR was estimated by Drakeley et al.,⁹ indicating a stable mesoendemic MTI level (PfPR = 10–50%) in those areas.^10,16 Plasmodium falciparum parasite prevalence varied largely between sites as it was expected from various preselected transmission intensity areas. In both surveys, the two sites in BF had the highest PfPR and the two sites in SN, the lowest. Plasmodium falciparum parasite prevalence rates recorded in this study are in line with previous findings.^9,16–18

Across most sites, the PfPR was lower in the 6-month to < 5-year than in the 5-year to < 10-year age-group. In addition, a risk factor analysis highlighted an association between older age and higher risk of infection. Those results corroborate the findings of other studies, identifying increasing age as a risk factor for carrying malaria blood stage parasites.^9,16,19 This may potentially be explained by the fact that younger children benefit from a more focused usage of control interventions (bednets).²⁰ Another explanation would be increased immunity in older children due to repeated exposure to the parasite, leading to asymptomatic carriage and a lower probability to be treated than in symptomatic children.^21,22

Infection with P. malariae was observed in a few participants, and infection with P. ovale was rare, which supports observations from other studies conducted across SSA.^23–26 Across all sites and surveys, only one Plasmodium vivax infection was identified by microscopy in TZ, which differs from Twohig et al.’s²⁷ recent findings of growing evidence of this species in SSA. A high proportion of coinfections with P. falciparum was observed in P. malariae–infected participants (between 70% and 90%) and P. ovale–infected participants (around 60%). Similarly, high percentages of P. falciparum coinfection in P. malariae–infected participants was previously reported in Guinea (97%),²³ Uganda (91%),²⁴ and in the Democratic Republic of the Congo (90%),²⁵ with lower percentages reported in rural BF (67%)²⁶ and Benin (34%).²³

Approximately 20% of participants positive for P. falciparum by QT-PCR were undetected by microscopy. Classifying the study sites by MTI, this proportion was higher in low MTI sites. This might be explained by a relatively lower sensitivity of microscopy readings than nucleic acid related techniques, particularly in low parasite density infections, which are more frequently observed in low MTI areas.^12,13 Nevertheless, the kappa statistic estimated a substantial agreement between microscopy and NAAT techniques using either qualitative or semi-quantitative real-time PCR.

Monitoring year-to-year variations in MTI and in the use of malaria control interventions.

In addition, the present results are establishing a standardized baseline for further estimation of the year-to-year variations in PfPR and in the use of malaria control interventions, which are key variables to be monitored before and after vaccine introduction in the MVIP areas. Little variation in P. falciparum prevalence between the first two annual surveys was observed among most sites. Usage of bednets as a malaria control intervention was high in all sites and in both surveys (ranging 70–99%). Some variation in the usage of bednets was observed in all sites, except in BF.

In S2, participants not using mosquito coils, insecticide sprays, or repellents against malaria vectors were significantly less likely to be infected with P. falciparum (OR: S2 0.88 [95% CI: 0.782; 0.988]); however, no such association was observed in S1. Various entomological studies have questioned the efficacy of repellents and coils as effective malaria prevention measures and highlighted the false sense of protection perceived by the user.^28–30 Our data do not allow to draw robust conclusions at this stage, and the trend observed in S2 should be closely monitored in the subsequent cross-sectional surveys.

Study limitations.

As with all interview–questionnaire-based studies, this study could have been subject to information or recall bias as data related to bednet usage, control interventions usage, and care-seeking behaviors were collected from parents’ recollection as opposed to objective observation. The impact of these potential information biases on the study results is estimated to be limited because of the proximity in time between the occurrence of the event for which information is collected and the interview itself. Another study limitation may be related to both sensitivity and specificity of microscopy slide readings. Slide reading performances may indeed vary depending on the parasite density and species identification. However, the kappa statistic estimated a substantial agreement between microscopy and QT-PCR, the latter being able to detect a higher number of low density infections than microscopy and RDT.^12,31 Moreover, this should mainly impact low transmission settings due to the high proportion of low-density infections.