Within the TFM concession, SMPS was used as the primary measurement tool for establishing a pre-control baseline malaria prevalence rate and subsequently the basis and means of quantitatively assessing the impact of the MVCP interventions over time. For intervention period surveys, schools were randomly selected in a proportional manner per HA to more accuracy represent the prevalence of malaria in the various coverage areas (urbanized and rural) inside the FHZ. By providing comparable longitudinal data from a reliable and informative population cohort, this evidence has been used to assist in the timely review and, when necessary, the reorientation in control programme policy and operational activities.
Standardized prevalence surveys were found useful as an objective, evidence-based tool for detection of increased malaria activity that was associated with the rapid development of insecticide resistance in the primary Anopheles vector mosquito population to the pyrethroid class compounds being used for indoor residual spraying (IRS) and factory-treated bed nets. This lead to the use of an alternative class compound (carbamate) for IRS (Fig. 2). Routine SMPS detected a more persistent high malaria prevalence in many rural health areas compared to more urbanized locations, indicating that additional or more intense interventions are necessary to reduce malaria burden in these more at-risk communities. Historically, longitudinal information of this type is very rare in the DRC, especially in areas far removed from the Kinshasa area where our control programme activities take place [30,31,32,33,34].
The baseline SMPS in 2007 confirmed that malaria was the leading cause of morbidity in the FHZ. Soon afterwards, a MVCP was set up with IRS as the main pillar for vector control in the community. Following the initial rounds of house-to-house indoor spraying, the mean malaria prevalence decreased significantly from above 77 to 22.6% by October 2009. Afterwards the biannual prevalence has varied from a wet season high of 47% (May 2011) to a dry season low of 21.7%. The prevalence at the end of this intervention reporting interval (October 2014) was 33.5%, an overall 55.8% reduction in malaria since control programme inception.
Following the introduction of broad coverage vector control measures, the malaria prevalence decreased rapidly. This result that can be attributed to the first-time introduction of pyrethroid-based IRS in the majority (> 90%) of households combined with mass distribution commercial long-lasting insecticide-treated nets. The pyrethroid active ingredients used (deltamethrin, lambda-cyhalothrin, and alpha-cypermethrin, as either wettable powder or granule formulations) provided +an approximate effective residual life of between 4 and 6 months depending on the type of surface sprayed. Initially, the single annual IRS spray round began just before the rains resumed with the intention of providing protection during the higher malaria transmission period (October to March) of the year. The survey strategy was based on the expectation for natural reduction in transmission between May and October attributed to negative effects on the local vector population during the prolonged dry period.
The initial steep reduction was seen up to the October 2010 (22.6%) survey, approximately 2 years following IRS introduction. However, from October 2010 onwards, the prevalence increased significantly (p < 0.001) from 30.6% up to 47% in May 2011. Beginning in May 2010, the MVCP entomology laboratory and insectarium was established and two insecticide susceptibility bioassays were performed: (1) the standard WHO tube contact test using wild-caught larvae reared to the female adults and exposed to insecticide-treated filter-paper according to protocol [35] and (2) field-based cone bioassays [36] on treated walls of homes to determine the sensitivity and longevity of insecticides on various types of sprayed surfaces against both a colonized insecticide-susceptible strain of An. arabiensis and wild-caught anophelines, primarily An. gambiae s.s. ‘S’ molecular form. Results from 2010 susceptibility assays found conclusive evidence of substantial levels of resistance in An. gambiae to pyrethroids (permethrin and deltamethrin) including cross-resistance against DDT. These tests confirmed that in a period of 2 years from inception of wide scale IRS and LLINs (both using pyrethroids), selection pressure was sufficient to produce operationally relevant resistance (> 20%) against pyrethroids. Thus, in 2011, a change was made from pyrethroid to a carbamate-based insecticide (bendiocarb). As carbamates typically have a shorter residual life of only 3–4 months on most sprayed surfaces, two spray rounds were required to cover the entire high malaria transmission period. In May 2012, a second mass distribution of LLINs was conducted. Together with the IRS, treated bed nets likely contributed to the reduction of malaria from May 2011 to October 2012, a period which witnessed the lowest slide positive rate (21.6% based on adjusted RDT findings) over the 6-year observation (Fig. 2). In late 2014, with a planned rotation, bendiocarb was replaced with an organophosphate compound (pirimiphos-methyl) to mitigate the development of resistance. The decisions to change insecticides were predicted on data provided by regular SMPS alerting the programme on control performance indicators including routine insecticide susceptibility testing of the local vector population.
Although a far greater majority students without fever were found infected with malaria (96%) than the few who presented with fever at time of examination, there was a significant association between those with fever (≥ 37.5 °C axillary temperature) and having a malaria infection (41.8%, p = 0.002) in both survey periods of the year. Students with fever were 1.3 times as likely to be infected with malaria compared to the group without fever. This agrees with other studies showing a strong linear relationship between the percentage of febrile children with malaria and infection prevalence [37, 38]. When data were aggregated by survey location (urban and rural), there was also a significant association between presence of fever and malaria infection in urban localities (62% of fevers had malaria) while showing no such relationship in rural areas (only ~ 30% of fevers had concurrent malaria). It is surmised that children from rural areas might have a greater risk of febrile illness due to other causes and/or possibly a higher level of naturally acquired immunity to repeat malaria infection thereby suppressing febrile episodes. It is also a possibility that rural children might be more likely not to attend school on day of fever regardless of reasons (i.e., due to other demands rather than feeling ill such as assisting family in the field). These possibilities would require further investigation to determine which factors might be responsible for separating -urban and rural child populations.
Different age groups among school-aged children have been used to estimate the risk of malaria infection in an area. From examples in Uganda [39], Malawi [40], Mozambique [41], Equatorial Guinea [40], Côte d’Ivoire [42], Senegal [43], Tanzania [44], school-age children used for surveys were between 5 and 9 years of age. Other studies conducted in children in Uganda, Kenya, Tanzania, Mali, Nigeria, Central African Republic, Cameroon, Congo Brazzaville, Ethiopia, Somalia, Yemen and Ghana have selected ages from 0.5 up to 18 years [19, 20, 45,46,47,48,49,50]. A systematic review on studies from Africa published in a 20-year period defined ‘school-age children’ as those between 5 and 14 years [5]. The ages of students sampled ranged from 6 to 12 years. The DRC Ministry of Education defines the age of 6 as the legal age to begin the first year of primary school [51]. It was unusual in the context of these surveys to have children completing all primary school grades before 12 years of age. Six to 12 years of age is in line with most mathematical models linking malaria infections (transmission) with age of students [6, 8].
Malaria infection was aggregated in three age groups: 6–8, 9–10 and 11–12 years of age. Overall, malaria prevalence in these age groups were 32.8, 35.4 and 34.0%, respectively. Malaria infection was independent of age or gender and there was no significant association between presence of malaria infection and age groups. In Kinshasa, Kazadi et al. [33] examined school-based malaria surveys conducted between 1981 and 1983 and found no significant difference in parasite rates by sex in any age group (5–9, 10–14 and ≥ 15); whereas infection prevalence increased significantly (p < 0.001) with age. The authors attributed this infection discrepancy to younger children being more prone to becoming more symptomatic with malaria and therefore less likely to attend school those days when ill compared to older children. Thus, confounding caused by increasing levels of naturally acquired immunity (i.e., decreased likelihood of more severe morbidity) in the older children enabled a higher probability of them attending school despite infection. They also surmised the difference might be that younger children received antimalarial treatment more often as was also described in Brazzaville [52] to explain why older children had an apparent higher malaria prevalence. Both Kinshasa and Brazzaville are large urbanized areas and show little resemblance to the more rural-based, smaller population of the Tenke-Fungurume area. In rural Ghana, Sarpong et al. [20] found a strong variation of malaria prevalence across schools, but a declining risk of parasitaemia (mean parasite densities) with age; whereas the frequency of patent infections without fever was higher in the more remote villages. The general decline in malaria incidence as age increases has been documented in other areas of Africa [53]. For example, the slide positivity rate (SPR) has been used to estimate changes in malaria incidence (number of cases per person-time) in Ugandan children aged 1–10 years and only those presenting with fever at time of testing [54]. They reported that younger age was associated with significantly greater risk of malaria and even more so during the seasonal peaks of transmission. Thus, it was concluded the SPR could be used as a surrogate measure in routine surveillance to describe changes in burden of malaria and incidence between seasonal variation and different age groups.
In this current 6-year retrospective analysis of school-based prevalence in DRC, no significant differences in malaria infection between age groups in the sampled population was seen. This may indicate that twice-yearly surveys may not be sensitive enough monitoring to detect a difference, if present. The SPR can be a useful measure of the impact of control interventions when considering several important caveats stressed by Jensen et al. [54]. First, a change in SPR does not necessarily equate to a proportional or linear change in actual incidence; second, age and seasonality can affect incidence and, therefore, may distort the actual impact of interventions and third, although the SPR can be used to estimate relative temporal-spatial changes in malaria incidence, it is not an estimate of actual incidence in the population. Unlike the Ugandan study, in the DRC analysis, children were randomly selected from the general population, regardless of fever status at time of sampling, thus the SPR was not directly affected by any change in incidence in non-malarial fevers that would potentially result in a change in SPR by selection bias and not reflective of true change in malaria incidence.
Findings showed no difference in malaria prevalence by gender. Other than apparent equal exposure to malaria risk as there would be no apparent occupational-related risk factor in most children at or below 12 years of age. With this younger age group, parents would be as likely to provide the same level of health care to offspring regardless of gender. Female gender inequality, when present, is generally observed into later adolescence and adulthood [55, 56].
In the SMPS, the overall risk of being infected by location was 1.92 greater in rural compared to urbanized localities, regardless of time of year. These results indicate a strong link between malaria transmission and location. A meta-analysis of entomologic inoculation rates (EIR) from urban, peri-urban and rural studies published between 1977 and 2000 reported a mean annual EIRs of 7.1 in city centers, 45.8 in peri-urban areas and 167.7 in rural areas of sub-Sahara African countries sampled [57]. Other studies made the similar conclusion that malaria is far more prevalent in rural compared to urban areas. In rural settings, the proximity to permanent larval habitats sites is probably the aggravating factor [58]. However, rural environments have also indirect links to increased malaria risk through several factors including lower health spending and more limited social-health resources [59]. In urban settings, although malaria transmission could be low due to the urbanization/environmental changes itself, other factors like pollution could adversely affect the suitability of larval habitats, mosquito survival and ability or likelihood to transmit malaria (i.e., vectorial capacity). Other differences might include better housing with physical barriers such as screens, doors, use of insecticides and bed nets. Higher human population densities may also reduce individual biting rates (thus risk), owing to the higher ratio of humans to mosquitoes [60, 61].
The distribution of malaria infections was significantly different between rural and urban HAs and between May and October surveys, thus indicating an interaction between time of year and location. In May 2012, while the overall malaria prevalence in the concession decreased, there was a significant increase in malaria in rural localities and a decrease in urban localities. This suggested that additional interventions to the existing methods to control malaria (IRS and LLINs) were required in rural localities to reduce the malaria burden. In response, a pilot study on community-based malaria diagnostic and treatment was conducted from 2012 to 2013 in 14 rural villages. Results indicated that a scaled-up community-based approach using trained and supervised community health workers could be an effective and promising additional strategy to control malaria and an addition to the integrated MVCP [62].
The prevalence of single infections seen with P. falciparum, P. malariae and P. ovale were 29.9, 1.8 and 0.3%, respectively. P. malariae and P. ovale where consistently detected at low rates compared to P. falciparum, and typically at low density. In another study, Congolese children under the age of five showed a higher prevalence of P. malariae and P. ovale, 12.9 and 8.3%, respectively [63]. The variation is likely attributed to the difference of the age groups sampled (under 5 versus 6–12 years of age in this 6-year TFM retrospective analysis) as well as the diagnostic methods used (PCR versus expert microscopy) [34].
Overall, there was a significant association, albeit weak, between P. falciparum density and different age groups (p = 0.039). Additionally, there was an inverse correlation between survey period and P. falciparum densities. However, in rural localities, the correlation was not different to zero while it was significantly different in urban localities, i.e. during intervention the mean P. falciparum density decreased over time in urban localities while showing no changes in rural localities.
Lateral flow immunochromatographic parasite protein-detecting assays (known more commonly as rapid diagnostic tests) can provide generally accurate diagnosis for case management, often with a similar analytical accuracy to microscopy when using a high-quality product, in good condition, and prepared and interpreted correctly by a trained operator. RDTs are more amenable for use in remote locations and less dependent on higher skill levels to perform [64]. Although RDTs are a major advancement to malaria control, there are several limitations present in the current products on the market and depending on test detection format. For example, one of the most significant drawbacks is reduced test sensitivity when parasite densities are low [65], compared to other methods (e.g., expert microscopy, PCR). Therefore, to enable capturing a greater number of lower parasite density infections in our sampled population, we took the additional step of having matched blood films carefully examined by expert microscopy.
Typically, PfHRP2-detecting RDTs outperform (better sensitivity) than non-PfHRP2 assays that use either aldolase or Plasmodium lactate dehydrogenase (pLDH) enzymes for detecting infections. As P. falciparum is overwhelmingly the predominant malaria parasite species seen in the FHZ, we deliberately selected high quality PfHRP2 detection RDTs. Greater thermal stability compared to pLDH-based RDTs and product availability also influenced the choice of product [66].
The recognition of PfHRP2 and PfHRP3 parasite gene deletions that can result in false negative RDTs [67] should be assessed at TFM site. Currently, based on the reported low frequency of this genetic anomaly in Africa and the low percentage of false negative HRP2 RDT results identified in this observation with infections showing moderate to high density P. falciparum parasitaemias, this is not view as a substantial issue, if any, in the FHZ at this time. If later deemed a problem, it will raise the need for using tests that use non-HRP2 falciparum-specific targets, i.e., pLDH, either pan (all Plasmodium species) or P. falciparum-specific, alone or in combination with HRP2. In future, those cases presenting with high parasitaemias that produce negative RDT results could be tested for possible described gene deletions to explain false negative results. Additionally, hyper-parasitaemia can result in an apparent ‘prozone’ effect due to an antigen overload resulting in a false negative result [68]. Prozone is also regarded an infrequent occurrence but appears a common limitation to many HRP2-based RDTs at frequencies that may diminish diagnostic accuracy. Conversely, PfHRP2 has been shown to persist and can be detectable many days, even weeks after the clinical symptoms of malaria have disappeared and the parasites have apparently been cleared from the host [69]. This would lead to a lower RDT specificity (false positive signal). However, low-level parasitaemias more common areas of constant exposure to malaria parasites may result in positive findings due to persistent circulating HRP2 from previously cleared parasitaemias and not necessarily because of a current infection. Moreover, one other limitation to HRP2-based RDTs is the weak relationship between HRP2 quantitative blood level and the peripheral parasite density [65]. HRP2-based RDTs do not detect the parasites per se but rather the concentration of protein analyte in circulation. Thus, it is not possible to translate amounts of target antigen produced by parasites into percentage of parasitized red blood cells or parasites per µl of blood, two commonly used thresholds describing the degree of infection (i.e., severity). Therefore, comparing RDT performance sensitivity with microscopy results can present interpretive hurdles. Nevertheless, the addition step of using light microscopy to quantitatively measure parasite densities provided valuable information on malaria in the intervention area that would not be available using an RDT alone.
In the school-aged population surveyed, the RDT sensitivity (false negative) has been a notable limitation as the vast majority of such cases appear to have been the result of low-level parasitaemias/asymptomatic infections (generally densities below 100 asexual parasites/µl peripheral blood). This would be of less clinical significance compared to having greater epidemiological implications as repeat infections due to regular exposure are more likely to become subclinical parasite reservoirs.
Many contributing factors and limitations in data collection are acknowledged in the SMPS presented, but many were beyond the scope, practicality, and intention of this routine monitoring exercise. For example, time and resources were not available to follow-up registered children not attending school the day of sampling. Children could have been absent for many reasons, one possibility being illness caused by malaria. The loss of children with malaria would have biased the findings and analysis on factors such as age and fever. However, to what extent is not known. Different school’s attendance rolls the day of survey were not compared to determine percent absence and if differences were significant between locations (rural vs. urban, for example). Depending on location, local demographics may vary considerably, thus a contributing for likelihood of school attendance. For example, in rural areas, more dependent on subsistence seasonal agriculture, a child’s absence may be the result of having to assist in the field during peak planting or harvesting periods. On the other hand, school absenteeism in the urban area could be due to greater access or more frequent travel outside the area. Lastly, it would have been preferred that two persons independently examine the blood films using light microscopy. However, this was deemed prohibitively expensive and time consuming. Moreover, no matter the number of independent reads, some very low density/sub-patent infections will go undetected nevertheless. Given the level of expertise of the microscopist used for all 11 surveys, the authors are confident the results are accurate enough and comparable across surveys for the intended purposes described herein. Lastly, fever (≥ 37.5 °C) was measured using a digital device placed in the axillary area. Axillary temperatures are usually 0.3–0.6 °C lower than an oral temperature. An axillary temperature is influenced by the temperature on the outer surface of the body. A normal axillary temperature is between 35.9 and 36.7 °C. In other words, ‘fever’ during sampling was likely underestimated. While oral readings are typically more accurate, they can take longer to measure than axillary temperatures and thus impede mass sampling efforts. Although the overall absolute number of fevers may have been lower using the axillary method, as all children were measured in the same way, this should have had no meaningful impact on the statistical comparisons.
The surveys and findings were solely derived for operational purposes, not as controlled experiments, and thus output has some inherent constraints. The school malaria observations were made as part of an evidenced-based monitoring strategy to provide important metrics (i.e., disease prevalence) to directly assist the control programme. What is presented is the first known longitudinal set of data on malaria prevalence in school children in a defined, but rapidly expanding, population in southern DRC. To this point, recent information on the malaria burden in rural areas of the Katanga region (formerly Katanga Province, now sub-divided into four separate provinces), has remained fragmented, anecdotal, or non-existent in most instances. Since the cessation of organized malaria control programmes in the area in the 1980s mainly by state-sponsored companies such as the mining operations (Gécamines) and the DRC Railway Company (Société Nationale de Chemin de fer du Congo-SNCC), there has remained a large gap in basic epidemiological data and information on malaria in the region [70].
Effective malaria control and planning requires accurate measurements and information on both the geographical distribution of malaria risk and the effectiveness of malaria interventions [16, 71,72,73]. School-based surveys of children can provide a rapid, sensitive, convenient, and sustainable approach to malaria disease monitoring and evaluation and have been recommended as a complement to passive heath care statistics and household surveys [4, 5, 16, 19, 20, 74]. By contrast, household surveys, although informative, are generally more expensive, time and labor intensive and are more typically undertaken every 3–5 years or at longer time intervals [16, 74]. Advantageously, school-based surveys can be conducted more frequently, while also concentrating on an age group more likely to reside in the area of interest the majority of the time. Unlike older adolescents and adults who may travel out of the area more frequently and thus have a different malaria exposure risk profile. Pre-school children (< 60 months) would also be more likely to represent local exposure, however this age group poses two distinct disadvantages—typically a less convenient cohort to access that often requires door-to-door visits, and lower sample representation in a population. In the FHZ, the biannual SMPS will continue to be performed to monitor impact of the malaria control programme, the effects of interventions by location and time. This approach has provided important operational performance indicators to assist in programme decisions on control prioritization, use of insecticides, and timing of interventions. In operational control programme settings, longitudinal epidemiological measures are needed to measure the changing landscape of malaria risk over time. Better utilization of school-based health surveys (not just malaria) in ongoing surveillance programmes, support of health management information systems, and identifying most prudent use of current control methodologies should be promoted.