Ethics Statement

The study was approved by the Kenya Medical Research Institute and National Ethics Review Committee (SSC number 1543), the London School of Hygiene & Tropical Medicine Ethics Committee (5503), and the Harvard University Committee on the Use of Human Subjects in Research (F17578-101). Prior to the randomisation, meetings were held with community and school leaders and parents/guardians in each school to explain the study objectives and procedures. Parents/guardians of all children in classes 1 and 5 were requested to provide individual written informed consent and they were given the option to withdraw their child from the study at any time. Prior to every IST round or assessment, the procedures were explained to the children and they were required to provide verbal assent. An independent data monitoring committee reviewed the trial protocol, data analysis plan, and preliminary results.

Study Area and Population

The trial was conducted from January 2010 to March 2012 in Kwale and Msambweni districts on the south Kenyan coast (Figure 1). Malaria transmission in the area is moderate and perennial with seasonal peaks following the two rainy seasons (April–July and September–November) [31]. The primary malaria vectors are Anopheles gambiae s.l. and A. funestus [32],[33]. Intensity of malaria transmission has been declining in recent years: school surveys conducted in 2010 reported prevalences of P. falciparum of 9%–24% [34],[35], compared to 64% in 1998 [32]. Overall reported net use in the region is high, with the communities having benefited from universal coverage campaigns. During the two-year trial period, albendazole was delivered through households as part of the national lymphatic filariasis campaign in 2011, although coverage was not extensive and praziquantel was delivered to schools in the area in June 2011. The vast majority of the population in these districts belong to the Mijikenda ethnic group, with Digo and Duruma the predominant subgroups [36]. The region is primarily rural with subsistence farming of maize and cassava practiced by many of the communities, although titanium mining has recently become an important source of employment. In economic and educational terms, the districts are ranked the seventh poorest in Kenya and consistently have some of the worst performing schools in the national school examinations [37].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Map of the study area and schools.

Schools assigned to the IST intervention are shown in blue and schools assigned to the control group are shown in yellow. Insert shows the location of the study site in Kenya.

https://doi.org/10.1371/journal.pmed.1001594.g001

Kwale District has 85 schools across four zones, and in two of these an alternative literacy intervention study was underway. Therefore only 20 schools from Mkongani and Shimba Hills zones were included in our study, allowing the two interventions to proceed without leakage. In Msambweni District 81 of the 112 schools were selected, with schools in Lunga Lunga and Mwereni zones greater than 70 km away from the project office excluded because of logistical considerations in visiting them.

Study Design

The study was designed as a factorial cluster randomised trial to investigate the impact of two interventions: (i) the impact of school-based IST for malaria on the health, sustained attention, and education of school children, and (ii) the impact of a literacy intervention on education. In order to evaluate the potential interaction between the two interventions, schools were randomised to one of four groups, receiving either: (i) IST alone; (ii) the literacy intervention alone; (iii) both interventions combined; or (iv) control group where neither intervention was implemented (Figure 2). The study was not blinded. Because of the factorial design of the trial and lack of interaction detected (interaction effect p-values of 0.45, 0.26, and 0.60 for the three key literacy outcomes) between the two interventions in class 1 where both were implemented, we report the results of the interventions separately. Only the IST intervention results are reported in this paper. The results of the literacy intervention will be reported in a separate paper targeting an education research audience as the literacy intervention was focused purely on enhanced English and Swahili literacy instruction and was not intended to have an impact on health.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Study design diagram.

This figure depicts the randomisation procedures.

https://doi.org/10.1371/journal.pmed.1001594.g002

Recruitment and baseline sample collection were conducted in January–March 2010 using children randomly selected from classes 1 (age range: 5–15 years) and 5 (age range: 8–20 years). Both classes received the IST intervention, but the literacy intervention was targeted only to children in class 1 and as they advanced to class 2, as it focused on the initial stages of literacy acquisition. Education outcome measures were assessed in the same children at 9 and 24 months and health outcome measures at 12 and 24 months. Full details of the eligibility, randomisation, intervention procedures, and baseline results have been presented elsewhere [30],[35]. The study is registered with ClinicalTrials.gov, NCT00878007.

Sample Size

The sample size was based on methods designed for cluster randomised trials and assumed that 101 eligible schools would be randomised to the four intervention groups, with an average of 50 children per school. On the basis of data collected previously in the study area, the baseline prevalence of anaemia was assumed to be 20% and the coefficient of variation (CV) 0.2 [30]. In order to detect a 25% reduction in the prevalence of anaemia between the two groups, based on previous work in Kenya [24], the sample size required to give a study with a power of 80% at a two-sided significance level of 5%, was a total of 27 schools in each arm with 50 children per school. A sample size of 101 schools with 25 children per class (i.e., analysing classes 1 and 5 separately) will enable us to detect, with 80% power and 5% significance, an approximate difference of 0.2 standard deviations (SDs) between arms of the trial in educational achievement (assuming an intraclass correlation coefficient [ICC] of 0.2 and a pre-post correlation of 0.7), and a difference of approximately 0.15 SD in tests of sustained attention (assuming an ICC of 0.1 and a pre post correlation of 0.7) [24]. The increased number of schools required for the sustained attention and educational achievement outcomes provided greater power (97%) to detect a 25% reduction in the prevalence of anaemia, or alternatively 85% power to detect a 20% reduction.

Randomisation

The 101 schools were randomised in two stages (Figure 2), with each stage conducted during a public ceremony. In Kenya, schools are aggregated into sets of between three and six closely located schools, which regularly meet and share information, supported by a Ministry of Education Teacher Advisory Centre tutor. Our 101 study schools formed 24 of these sets of schools, which were randomised either to receive the literacy intervention or to serve as the literacy control. Randomisation of these sets of schools was stratified by (i) set size, to ensure equal numbers of schools in the experimental groups, and (ii) average primary school leaving exam scores of the school sets, to balance the two study groups for school achievement. This randomisation procedure was designed to minimize contamination of the literacy intervention methods across the study groups. In stage two, the IST intervention was randomly allocated at the level of the school, with the 101 schools re-stratified by (i) literacy intervention group assignment and (ii) quintiles of average school exam scores, producing ten strata overall.

Enrolment

At enrolment, children's height and weight were measured, axillary temperature was digitally recorded, and finger-prick blood samples were obtained to prepare thin and thick blood films and to determine haemoglobin concentration (Hb). Children known or suspected to be homozygous for sickle cell trait or pregnant were excluded. Any child found with Hb<80 g/l was referred by the nurse to the nearest health facility for iron therapy, and any child found with Hb<50 g/l was taken to the hospital for transfusion. Baseline parasitaemia was measured in the intervention group during the first round of screening but was not measured in the control group owing to the ethical constraints of testing for malaria but not treating children found to be infected in the control schools, which was of particular importance at baseline as the intervention involved screening of Plasmodium infection.

A questionnaire was administered to parents/guardians to record information on residence, family size, ownership of possessions, mosquito net use by them and their children, recent deworming of the child, house construction, and parental education level.

Intervention

IST was outlined as a possible strategy in the “Malaria-Free Schools Initiative,” as part of the Kenya National Malaria Strategy 2009–2017 [38]. During IST, children were screened once a school term for malaria parasitaemia using an RDT (ParaCheck-Pf device, Orchid Biomedical Systems), which is able to detect P. falciparum. Screening was conducted by laboratory technicians. Repeat visits were made to follow-up children absent on the day of screening. Children (with or without malaria symptoms) found to be RDT-positive were treated with a six dose regimen of AL (artemether 20 mg/lumefantrine 120 mg, Coartem, Novartis) over three days. Doses of AL were based on weight, with children stratified into one of the four categories (<15 kg, 15–24.9 kg, 25–34.9 kg, and ≥35 kg). AL was given at a dose of 20/120 mg to children <15 kg, 40/240 mg to children 15–24.9 kg, 60/360 mg to children 25–34.9 kg, and 80/480 mg to those who weighed ≥35 kg. Parents or older siblings of children were called and a nurse explained that their child was infected with malaria parasites and required treatment. Doses 1, 3, and 5 were given under direct observation at the school by the study nurses. Children were given milk and biscuits with the AL and observed for 30 minutes after drug administration. If vomiting occurred during this period, drugs were re-administered. If vomiting occurred on a second occasion, this was noted but the drugs were not given again. Such children were not excluded from the trial and they were eligible to receive drugs on the subsequent two days. The parents/older siblings or study children themselves if in the older classes were given doses 2, 4, and 6 each day for evening administration and provided with instructions on treatment. Children absent from school on days two or three of treatment were followed up at their home by the nurse, and provided with the doses. Supervised treatment was defined as nurses administering and directly observing doses 1, 3, and 5 taken on three consecutive mornings in the school and recording doses 2 and 4 reported by the child as having been taken the previous evenings. No direct confirmation of whether dose 6 was taken was recorded by the nurse. The record of supervised treatment was used as a proxy for compliance. Five rounds of screening and treatment were implemented. The first round was conducted alongside baseline health assessments in March 2010, the second round in July 2010, the third in September 2010, the fourth in March 2011, and the final round in October 2011.

Adverse events were monitored by the study team for 24 hours after each treatment, and a further 28 days thereafter using a passive surveillance system in schools. Travel costs were reimbursed and treatment charges waived. Adverse experiences were monitored until the event was cured or had stabilised. Agranulocytosis and hepatotoxicity were not assessed because of logistical constraints.

Follow-up

Cross sectional health surveys were carried out at 12 and 24 months. During these surveys, temperature, weight, and height were measured and a finger-prick blood sample was collected for determination of malaria parasitaemia and Hb. Children with an axillary temperature ≥37.5°C were tested using an RDT, providing an on-the-spot diagnosis for malaria, and treatment was administered as per national guidelines.

Laboratory Methods

Hb was measured using a portable haemoglobinometer (Hemocue). Thick and thin blood films were stained with Giemsa, asexual parasites were counted against 200 white blood cells (WBCs), and parasite density was estimated assuming an average WBC count of 8,000 cells/µl. A smear was considered negative after reviewing 100 high-powered fields. Thin blood smears were reviewed for species identification. All blood slides were read independently by two microscopists who were blinded to group allocation. Discrepant results were resolved by a third microscopist.

Attention and Educational Achievement

Tests of sustained attention and educational achievement were administered at baseline, 9 months, and 24 months. Sustained attention was a primary outcome, assessed through the code transmission test, adapted from the TEA-Ch (Tests of everyday attention for children) battery [39]. A recorded list of digits is read aloud and children are required to listen for a code—two consecutive occurrences of the number 5—and then record the number(s) that preceded the code. To avoid floor effects, (in which the assessment is too challenging to establish the range of abilities in the target population), a simpler measure of sustained attention—the pencil tap test [40] —was used at baseline for the younger cohort. Children were required to tap a pencil on the desk a predetermined number of times in response to the assessor's taps. The secondary outcome of educational achievement was measured through tests of literacy and numeracy. Literacy was assessed through group administered English spelling tests, adapted from PALS (Phonological Awareness Literacy Screening) [41], with the younger classes asked to spell five three-letter words and credit given for phonetically acceptable choices for each letter and the older classes asked to spell 25 words with credit given for correctly spelling the features and sound combinations of the word. Numeracy assessments involved an oral test of basic arithmetic for younger children at baseline and 9-month follow-up and written arithmetic at 24-month follow-up and a written arithmetic test throughout for older children. All educational assessments were piloted prior to use in the baseline and follow-up evaluations. During piloting the assessments were conducted under the same assessment conditions on two occasions a week apart, with the correlation between the scores at the two time points providing a reliability score. The inclusion criteria for the tests used in this trial was a Cronbach's alpha correlation of 0.7 or above, indicating a well constructed test with consistent administration.

The educational assessments were conducted separately to the health assessment both for logistical reasons and so as not to cause bias during the educational assessments due to apprehension of the finger-prick. The education assessments preceded the latter by an average of a week at baseline and 24-month follow-up. However, during the first follow-up, the education assessments were conducted at the end of the school year (9 months) and the health assessments were conducted at end of a full year (12 months).

Data Analysis

Data were double-entered, consistency checks were performed, and all analysis was conducted using Stata software version 12.1. The pre-specified primary outcome measures were the prevalence of anaemia, defined according to age and sex corrected World Health Organization (WHO) thresholds: Hb<110 g/l in children under 5 years; <115 g/l in children 5 to 11 years; <120 g/l in females 12 years and over and males 12 to 15 years old; and <130 g/l in males over 15 years, with no adjustment made for altitude [42] and sustained attention. The pre-specified secondary outcomes were the prevalence of P. falciparum and scores for spelling and arithmetic. Reported information on ownership of household assets and household construction was used to construct wealth indices using principal component analysis [43] and resulting scores were divided into quintiles. Anthropometric measurements were processed using the WHO Anthroplus Stata macro [44] to derive indicators of stunting, thinness, and underweight.

The analyses described here correspond to a pre-specified statistical analysis plan, approved by both the data monitoring committee and trial steering committee before any data were examined.

Baseline school and child characteristics, together with baseline measurements of the study outcomes, were summarized by study groups separately, with class-specific study outcomes reported separately by class. Counts and percentages were used for categorical variables. Means and standard deviations, or medians and the limits of the inter-quartile range (IQR), were reported for continuous variables. Coefficients of variation (CVs) for the binary (health) outcomes and intraclass correlation coefficients (ICCs) for the continuous (cognitive and education) outcomes were calculated from the baseline measures using appropriate formulae [45].

The effectiveness of the IST intervention was assessed using generalized estimating equations (GEE) with robust standard errors and an exchangeable correlation matrix to allow for clustering within schools. All main analyses used the intention-to-treat principle whereby children were analyzed in the intervention group that they were assigned to, even if the child moved schools or did not fully comply. The primary pre-specified analysis adjusted for age (as a continuous variable), sex, and the baseline measure of the outcome, except for baseline P. falciparum, which was not measured in the control schools. As randomisation of schools to the IST intervention was stratified on the basis of both literacy intervention assignment and school mean exam score (Figure 2), all adjusted analyses presented account for these two stratification factors. Data for classes 1 and 5 combined were used for the health outcome analyses. However, as different assessments were administered for classes 1 and 5 for the evaluation of attention (e.g., pencil tap for class 1 and code transmission for class 5), literacy, and numeracy outcomes, analyses were conducted for each class separately. Separate GEE analyses were conducted for the first and second follow-ups. No formal adjustment was made for multiple testing, therefore p-values should be interpreted with due caution. However, as specified in the statistical analysis plan, formal testing was restricted to two primary and three secondary pre-specified outcomes.

For comparison purposes, we also obtained estimates from an unadjusted model that did not adjust for baseline outcome measures, child characteristics, or study design (literacy group and mean school-exam score) and hence retained all study children assessed at follow-up regardless of whether they had baseline measures. Secondary analyses were conducted additionally adjusting for stunting, school-feeding programme, and socioeconomic status (SES) on top of the pre-specified variables. These additional adjustments had no notable impact on the effect estimates and are not presented.

In order to gain power and account for missing data, random effects models, using a likelihood-based approach, were fitted to the one-year and two-year follow-up data simultaneously (Tables S1, S2, S3, S4, S5, S6, S7; Text S1). Additional sensitivity analyses were conducted to examine intervention effects when children who had transferred from their original school were excluded from the analyses (Table S8).

Role of the Funding Source

The funders had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Trial Profile and Baseline Data

One hundred and one schools were randomised to one of the two study groups (Figure 3). In total, 7,337 children aged between 5 and 20 years (median: 10 years and IQR: 8–13 years) were randomly selected in January 2010 of which 5,772 (78.7%) parents consented, with no real differences found between groups in terms of percentage of parents refusing and not attending the meetings. Overall, 5,233 children were initially enrolled, of which 5,176 (98.9%) children were eligible for follow-up after the baseline assessments. Characteristics of the children included in each of the study groups are shown in Table 1. The numbers of children per school ranged from 18 to 58 but overall were well balanced between groups (control: median, 52; IQR, 50–54 and intervention: median, 53; IQR, 50–55). A difference in percentage of children unavailable for the baseline health surveys was observed between the groups with 5.1% and 10.1% unavailable in the control and intervention groups, respectively (Figure 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Trial profile.

The flow of children and clusters in the 50 control 51 IST intervention groups at all assessment points throughout the two-year study period. FU1 indicates follow-up 1 and FU2 indicates follow-up 2. Cluster size is presented as mean (SD) [min, max].

https://doi.org/10.1371/journal.pmed.1001594.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of 5,233 study children in the 50 control and 51 IST intervention schools.

https://doi.org/10.1371/journal.pmed.1001594.t001

Compliance with Screening and Treatment

During the 24 months of intervention, an average of 2,340 children (88.4% of eligible study children) in the 51 intervention schools were screened at each visit, of whom, on average 17.5% were RDT-positive (Table 2). Of the study children, 84.0% were screened at four or more IST rounds and 66.8% were screened at all five rounds. By the fifth screening round, 3.3% children were lost due to withdrawal or death and a further 17.7% of children were lost due to out-migration. The percentage of children RDT-positive at each screening ranged from 14.8% to 19.2%, with no distinct trend over time. Overall, 99.1% of RDT-positive results led to treatment across the five screening rounds and 92.6% of these were recorded as receiving the fully supervised six-dose treatment regime (Table 2). There was an apparent decline in full supervision (a proxy for compliance) with time, falling from 96.9% at the first round to 81.7% at the fifth round. RDT performance, examined against a “gold standard” of expert microscopy, revealed consistently high specificity, greater than 90% at all rounds, whereas sensitivity was more variable ranging from 68.7% to 94.6% across surveys, with higher sensitivity observed during the wet season compared to the dry season (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary information for 2,710 study children in the IST intervention group by screening round.

https://doi.org/10.1371/journal.pmed.1001594.t002

Follow-up

Of the 5,233 children enrolled initially, 4,446 (85.0%) were included in the 12-month follow-up health survey and 4,201 (80.3%) were included in the 24-month health survey (Figure 3). At 12 and 24 months, children lost to follow-up across both study arms were largely similar to children followed up (Tables S1 and S2), with slightly lower spelling scores in those children lost to follow-up across both groups and a higher proportion of children whose parents had no schooling in those lost to follow-up in the intervention schools. The prevalence of P. falciparum, in the intervention group, was lower in children lost to follow-up (8.6%) compared to those followed-up (13.6%) at both 12 and 24 months.

Overall, 4,656 (89.0%) of children were included in the 9-month follow-up education survey and 4,106 (78.5%) in the 24-month follow-up survey. Children unavailable for the follow-up educational surveys at 9 and 24 months were similar across the two study groups (Tables S4 and S5), with a slight imbalance in SES and parental education categories seen between children available and unavailable for the survey in the intervention group. Additionally, baseline prevalence of P. falciparum was lower in children lost to follow-up (9.1%) compared to those followed-up (13.3%) in the intervention arm.

As intention-to-treat analysis was performed, no adjustment was made for children transferring between schools and study groups at the follow-ups. Overall, 308 children were recorded as transferred by the end of the study. Of those, 46 (0.9%), 71 (1.8%), and 308 (5.9%) children were assessed in a different school from their initial enrolment school, at 9-month, 12-month, and 24-month follow-ups, respectively. Sensitivity analysis excluding these transfers resulted in no change in direction or magnitude of results (Table S8).

Effect of IST on Anaemia and P. falciparum Infection

At 12-months follow-up, 2,148 children in the control schools and 2,298 in the intervention schools provided a finger-prick blood sample for Hb assessment, and at 24 months 2,027 and 2,174 children provided finger-prick samples in the control and intervention groups, respectively. There was no significant difference in the prevalence of anaemia between children in the two groups at 12- or 24-month follow-ups (adjusted risk ratio [Adj.RR]: 1.03, 95% CI 0.93–1.13, p = 0.621 and Adj.RR: 1.00, 95% CI 0.90–1.11, p = 0.953), respectively (Table 3); the same was observed in relation to mean Hb. There was also no significant difference in the prevalence of P. falciparum between study groups at 12 or 24 months. Subgroup analysis of the impact of IST intervention on anaemia according to Plasmodium infection prevalence at baseline (using 12-month estimates for the control group as a proxy for baseline), demonstrated no differential impact by prevalence category (<5%, 5%–19%, and 20%+) at either follow-up. Similarly, no difference was seen when analysis was stratified, within the intervention group only, by numbers of treatments received across the study period (Tables S9 and S10).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Effect of the IST intervention at 12- and 24-months follow-up on health outcomes anaemia and P. falciparum prevalence for study children.

https://doi.org/10.1371/journal.pmed.1001594.t003

Effect of IST on Attention and Educational Achievement

At both 9- and 24-months follow-up, there was no statistical difference in mean scores for sustained attention between study groups in either class with adjusted mean difference (Adj.MD): −0.44, 95% CI −1.09 to 0.21, p = 0.180 and Adj.MD: 0.28, 95% CI −0.23 to 0.79, p = 0.283 for classes 1 and 5, respectively at the 24-month follow-up (Table 4). Similarly there was no significant difference between groups on scores for spelling in the older class at 9- and 24-month follow-ups (Adj.MD: −0.31, 95% CI −1.26 to 0.63, p = 0.515 and Adj.MD: 0.71, 95% CI −0.34 to 1.76, p = 0.183) nor for arithmetic at either follow-up (Table 5). However, at 9-months follow-up, children in the younger class in the intervention group had lower mean adjusted scores for the spelling task and the same trend was observed at 24 months (Adj.MD: −0.65, 95% CI −1.11 to −0.20, p = 0.005). Similarly at 24 months, in the younger class, children in the intervention group scored on average 0.60 points lower in the arithmetic assessments than children in the control group (Adj.MD: −0.60, 95% CI: −1.02 to −0.19, p = 0.005).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Effect of the IST intervention at 9- and 24-months follow-up on sustained attention outcomes for younger (class 1) and older (class 5) children.

https://doi.org/10.1371/journal.pmed.1001594.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Effect of the IST intervention at 9- and 24-months follow-up on educational achievement (spelling and arithmetic) outcomes for younger (class 1) and older (class 5) children.

https://doi.org/10.1371/journal.pmed.1001594.t005

Surveillance for Adverse Effects

Active surveillance found that 4.5% (92/2,030) children reported one or more adverse effects within 2 days of receiving treatment, including headache (68; 3.3%), stomach ache (38; 1.9%), dizziness (17; 0.8%), vomiting (7; 0.3%), and pruritis (10; 0.5%). During the 24 months of follow-up, 11 children died: five in the intervention group and six in the control group. Cause of death was investigated and included yellow fever, heart defect, leukaemia, drowning, trauma, pneumonia, and paediatric HIV. In the intervention group, none of these deaths occurred within 30 days of the screening and treatment and therefore were not attributed to the intervention.

Conclusion

In summary, our findings show there are no health or education benefits of implementing school-based IST with AL in a low to moderate transmission setting such as this study site, as a high proportion of children screened do not require treatment and those who do largely live in focal high transmission regions, where rapid re-infection occurs between screening rounds and results in no lasting gains from treatment. Nevertheless, our results do highlight a potential role for schools as screening platforms. School screenings using RDTs could provide an operationally efficient method to initially identify transmission hotspots for targeted community control [72]. School surveys have proved a useful platform for defining heterogeneities in Plasmodium transmission over large geographical areas in a more rapid and low cost manner than community surveys [73],[74]. The results from this study's screening rounds present a case for the use of schools in also depicting local transmission heterogeneities, which can be extrapolated to the local community [75] and aid in developing targeted community-wide comprehensive interventions, such as localised indoor residual screening and larviciding, with biennial school screenings used to monitor the success of these interventions. The use of schools in this way is a focus of our current research.