Discussion
This study showed that, among regions where the RTS,S/AS01 Phase III clinical trial was implemented,
PfPR differed greatly by age and site; nevertheless, a higher malaria prevalence in the 5–19 years age group was observed consistently in all sites. Burkina Faso was the country with the highest
PfPRs over the 3 years, while the lowest prevalence was consistently observed in the 2 Tanzanian sites and Lambaréné in Gabon. In children aged 6 months–4 years, the age group for which the highest morbidity and mortality due to malaria is observed [
23], a significant decrease in
PfPR was observed over time at 3 sites: Kombewa (Kenya), Lilongwe (Malawi) and Bagamayo (Tanzania). The use of anti-malarial treatment was associated with reduced odds of
P. falciparum infection for all sites.
The higher
PfPR observed in Burkina Faso is not surprising and is in line with previous reports [
24]. In Nouna, Burkina Faso, malaria prevalence in children under 5 was 57.7% [
24], similar to that observed in Nanoro in our study. The results also confirm a lower
PfPR in the younger 6 months–5 years age group compared with children over 5 years of age reported in Nouna [
24]. A decline of
PfPR over time was observed when the surveys at the same 3 time points were considered, in line with reports of a consistently decreased prevalence between 2009 and 2011, compared to 2000 [
25]. A high prevalence was also observed in Kombewa, Kenya, for which data for the last 2 surveys seems to be in line with previous reports in a similar timeframe and geographical setting for children 0–5 years [
26], although larger values were observed in our study for the other age groups. Among the sites in Ghana, lower
PfPR values in Agogo than in Kintampo were consistently documented in our study. This difference is not surprising, as marked heterogeneity of
P. falciparum infection was previously reported [
27].
The lower
P. falciparum prevalence noted in the Tanzanian centres is in agreement with recent estimations of
PfPR by rapid diagnostic test in a community survey [
28], and reports of a decline in malaria incidence in this country, following scale up of malaria control interventions [
29]. Lilongwe, Malawi was the only other site where a consistent decrease over time was observed in, both for the 6 months–5 years and the 5–19 years age groups. The estimated
PfPRs compared fairly well with the
P. falciparum infection prevalence detected microscopically from a study conducted in southern Malawi from 2012 to 2014, although larger values were reported in the same study when real-time polymerase chain reaction (PCR) was used for parasite detection [
30]. In Lambaréné, the estimated
PfPR was lower in the 6 months–5 years age group than in the 5–19 years one, which is consistent with previous observation of an increase in age of the malaria high-risk population in Gabon between 2008 and 2011 [
31].
Ultimately, the
PfPRs’ heterogeneity across the 8 study sites represents well the variability of transmission intensity in sub-Saharan Africa [
32,
33], and confirms that RTS,S/AS01 vaccine efficacy was evaluated at varying levels of MTI [
34].
Differences across centres were also observed in relation to the use of anti-malarials and control measures, but they are also likely to be associated with other extrinsic factors such as social, behavioural, economic and local policies. Previous anti-malarial treatment (in all surveys) and use of bed nets (in ≥ 2 surveys) was associated with a lower risk of infection for the majority of the sites. During the study, the proportion of participants having taken malaria treatment varied widely in centres with high malaria prevalence; the lowest recorded use of malaria treatment was in Nanoro, Burkina Faso (0.5–2.5% of participants) and the highest in Kintampo, Ghana (up to 21.9% of participants). Of note, to our knowledge, Seasonal Malaria Chemoprevention measures as recommended by the World Health Organization had not been implemented at the time the study ended in any of the study sites [
35]. This finding may be explained by a number of highly interrelated extrinsic environmental, social, and economic factors such as diagnosis confirmation and local treatment practices, care seeking behaviour, access to health facilities and availability of local anti-malarials [
36]. The discrepancy between infection prevalence and infrequent use of anti-malarials at some sites may be related to the immunity afforded by chronic infections resulting in asymptomatic parasite carriage, which have been widely described in
P. falciparum endemic areas, including Africa [
37].
Bed net use (on the night before the survey) was high in all sites throughout the 4 years and was associated with reduced risk of infection. Although recent data confirm the effectiveness of ITNs [
7], reports in the literature from the period covered by our study did not always evidence an association between bed net use and lower odds of infection at all sites. For instance, while a national distribution campaign in Burkina Faso led to an increase in the use of ITNs, the more frequent use of bed nets was not associated with a decline in parasitaemia in children under 5 years of age [
38], as was the case in The Gambia [
27]. In Malawi, use of ITN following a high coverage of malaria interventions was associated with protection against infection, but
P. falciparum prevalence remained high, especially in school-aged children [
30].
Increased use of ITNs and IRS has already been associated with a reduction in malaria burden through a population effect, and the fact that the results of these measures have varied across regions [
39‐
47] and by age group [
30] is mainly due to issues in their implementation, although differences in the dominant malaria vector species could also play a role [
48]. Of note, in this study, information on using malaria medication or control measures was based on questions such as sleeping under a bed net the night before the survey, taking anti-malarial medication during the past 14 days, or exposure to IRS during the past 12 months, which may induce recall biases or generalize usage of bed net based on the report for 1 night. Moreover, malaria control interventions may be targeted to areas with higher parasite prevalence and this, together with variations in coverage within an area, may confound any observations.
As expected, both anaemia and fever were associated with
P. falciparum-infection in this study. OR analyses confirmed
P. falciparum infection to be a significant risk factor for anaemia, with ORs ranging from 1.21 to 4.08 across all sites and surveys (notwithstanding the negative association found in Bagamoyo at survey 3), consistent with results from other studies [
49‐
51].
Overall, fever was associated with high parasite densities (≥ 10,000 parasites/μl). There was a trend towards a higher association between fever and parasite density in the youngest age group, but results varied widely across centres. The lower fever prevalence observed in Nanoro, the site with the highest overall
PfPR for all surveys and age groups, seems to be in line with previous findings in Burkina Faso, with 85–90% of parasite-positive individuals showing no fever [
25]. However, in sites with consistently low
PfPR for the 6 months–4 years age group (Lambaréné, Lilongwe, Korogwe, Kintampo), fever prevalence was also substantial for low to medium parasite densities (< 10,000 parasites/μl). These results show that a lower parasite density was associated with fever at sites where
PfPR (and therefore MTI) was lower, as children in regions with moderate-to-high levels of MTI may develop partial immunity to malaria more easily and at an early age, manifested through a decrease in symptoms [
37,
52]. Of note, a larger prevalence of fever, regardless of recorded parasitaemia, was observed in Kintampo compared with the other study sites; moreover, in the 6 months–4 years age group, fever was reported for 31.0–50.2% of non-infected children. However, this observation is in agreement with reports of a high incidence of non-malaria fever in children born between 2008 and 2011 in the area [
53].
A number of key factors contributed to the quality and reliability of results presented in this paper. Standard methodologies were used across centres to ensure comparability of results and to make certain that sampling was representative across catchment areas which participated in the Phase III trial. This allowed for a surplus in the number of randomly selected households in order to take into account non-response and to minimize sampling bias. However, the sampling followed a non-probability method, as the targeted population represented a convenience sample, and therefore, generalization of the results should be made with caution.
The study has several limitations. Four surveys were conducted only in 3 study sites, as the study ended in December 2013, along with the Phase III trial, and at that time, 5 centres had not conducted the fourth survey. For some centres, the first survey was carried out outside the peak transmission season.
P. falciparum prevalence was assessed microscopically, a method which underestimates the true level of infection when compared with molecular (PCR) detection [
28,
54]. The cut-offs used to define anaemia and severe anaemia were based on the haemoglobin levels at sea level as recommended by the World Health Organization, but were not adjusted for each site, and this could have impacted the estimation of anaemia prevalence.
Malaria control intervention may have a different effect and/or impact on malaria morbidity at different levels of MTI. Therefore, the assessment of MTI in parallel to efficacy trials for new malaria control interventions is essential for the generalization of trial results to other settings and to guide their implementation [
55‐
58]. When administered as a 4-dose schedule at months 0, 1, 2 and 20, the vaccine efficacy of RTS,S/AS01 against clinical malaria estimated in the Phase III trial was 39.0% (95% CI 34.3–43.3), over a median follow-up period of 48 months since first vaccination of children 5–17 months old [
59], up to the year 2014. However, vaccine efficacy varied greatly among study sites, although there was no evidence for a statistically significant interaction with MTI [
10]. Of note, ordering the sites by increasing
PfPR for the 4 study years provided a similar ranking as when using incidence of clinical malaria, measured in control infants 6–12 weeks of age at enrollment in the Phase III study during 12 months of follow-up [
10,
11], especially for the 6 months–4 years and 5–19 years age categories. For all 4 years in our study, the site with the highest
PfPR was Nanoro, for which the highest malaria incidence was observed in the Phase III trial, while the lowest values were documented for Korogwe in both studies. The order of the other sites also compared fairly between
PfPR and clinical incidence of malaria, suggesting that the control group could serve as a surrogate for relative malaria parasite transmission intensity. The data in our study contribute to an improved interpretation of site variations in the Phase III vaccine efficacy study of RTS,S/AS01.
Authors’ contributions
SAb, SAd, TA, STA, AA, DA, KPA, JTB, TC, DC, UDA, LD, DBED, SD, CDr, JFF, SG, BG, IH, OJ, PK, DL, BL, PL, JL, PM, CM, FM, LM, MGM, RM, BO, LO, SO, WO, SOA, JO, JYP, ERB, MS, MCT, MT, GA and TH contributed to the conception, design or planning of the study. SAd, TA, STA, AA, DA, KPA, JTB, UDA, LD, DBED, SD, CDr, VE, JFF, SG, IH, OJ, DL, PL, JL, PM, CM, FM, LM, MGM, RM, BO, LO, SO, WO, SOA, JO, ERB, MS, MCT, MT, GA, EA, DAt, HOB, BEB, EU and TH collected or assembled the data. SAb, DA KPA, DBED, CDr, VE, SG, IH, JL, CM, FM, RM, LO, WO, SOA, JO, JYP, ERB, GA, EA, DAt, EU and MS performed or supervised the analysis. S Ab, TA, STA, DA, KPA, TC, DC, DBED, CDr, VE, JFF, SG, BG, ICH, OJ, PK, BL, JL, CM, FM, LM, RM, BO, LO, WO, SOA, JO, JYP, ERB, MS, GA, BEB, DCK, EU and MT interpreted the results. All authors read and approved the final manuscript.
Acknowledgements
The authors thank the study participants and their families and the study team members at each site. The authors thank those listed below by study centre who contributed in various ways to the trial: Albert Schweitzer Hospital, Lambaréné, Gabon: the study nurses, the field workers, the remote data entry team and the laboratory technicians. Ifakara Health Institute, Bagamoyo, Tanzania: Rehema Dangwa, Ali Said Hamad, Beatrice James, Caroline Mavere, Ernest Mnkande and Kafuruki Shubis. Institut de Recherche en Science de la Santé, Nanoro, Burkina Faso: Karim Derra, Blaise Gnoumou, William Kabore, Adama Kazienga, Ashmed Nana, Sayouba Ouedraogo, Eli Rouamba, Hermann Sorgho, Innocent Valea and Sandrine Yara. KEMRI-Walter Reed Project, Kombewa, Kenya: the IQC team, the nursing department, the laboratory department, the accounts department, the clinical department and the purchasing department. Kintampo Health Research Centre, Kintampo, Ghana: Haruna Seidu, the KHRC staff & management, the KHRC ethics committee, the KHRC scientific community, the KHRC Data management Centre and the KHRC Clinical Laboratory. Kwame Nkrumah University of Science and Technology, Agogo (Agogo), Ghana: Collins Agyeman, Ofori Amoateng, Evans Antwi, Isaac Donko, Ali Idriss, Amos Kotey, Leticia Kunaa, Frank Prempeh, Solomon Sawakia and Collins Paa Yeboah. London School of Hygiene and Tropical Medicine, London, United Kingdom: Lynn Spencer. Walter Reed Army Institute of Research, Silver Spring US: Evelyn Angove and Sheeted Data. National Institute for Medical Research, Korogwe, Tanzania: Marthe Lemnge, Edwin Liheluka, Anangisye Malabeja, Mohamed Mapondela, Zeno Manjulungu and Bruno Mmbando. University of North Carolina Project, Lilongwe, Malawi: Alice Banda, Towera Banda, Beatrice Chafa, Chikondi Chapola, Francis Chasakala, Fortunate Chinunda, Portia Kamthunzi, Pluxidia Kanduku, Shiraz Khan, Cynthia Libale, Sibongile Mafuleka, Patience Mulewa, Tisungane Mvalo, Aubrey Mwantisi, Rutendo Nkomo, Dalitso Nyakuipa, Gift Nyasulu, Severiano Phakati, Hannah Stambuli, Gerald Tegha, Tapiwa Tembo, Olivia Yambeni and Agnes Zilore. PATH Malaria Vaccine Initiative, Washington DC, USA: Shannon Simpson. GSK, Wavre, Belgium: Sarah Benns (freelance on behalf of GSK) and Petronela M. Petrar (XPE Pharma and Science on behalf of GSK) for scientific writing, Myriam Wilbaux and William Zonta (XPE Pharma and Science on behalf of GSK) for publication management, Denis Sohy for publication coordination and critical review, Mattéa Orsini for critical review and Laurence De Waele, Marcela Gavigan, Itumeleng Siweya and Corinne Willame for their contribution to the study measures, by site.