A cluster randomized trial of delivery of intermittent preventive treatment of malaria in pregnancy at the community level in Burkina Faso

With an estimated population of 20 million, Burkina Faso is a high malaria endemic Sahelian country in the heart of West Africa accounting for 4% of the world’s annual malaria cases [14]. Women comprise 51.8% of the overall population, and it is expected that 5.5% of women will be pregnant at any given time [15]. The study was implemented in three of the 13 regions (Sud-Ouest [South-West], Centre-Sud [Central-South], and Centre-Est [Central-East]) with highest malaria transmission. One district was purposively selected from each of the three regions, taking into account malaria epidemiology, IPTp coverage, and the presence of active CHWs (Fig. 1). Routine HMIS data indicate that before the trial, IPTp3 coverage in the selected districts, Batié, Pô, and Ouargaye, was 34.6%, 52.8%, and 47.2%, respectively, for an overall average of 44.9%.

Non-contiguous health facilities (HF) in each district were matched by estimated general population and estimated number of pregnant women and IPTp2 coverage (based on routine health facility data). In each district, two pairs of HFs were purposively selected, and one HF in each pair was randomly assigned to the intervention, for a total of 12 HFs (six intervention, six control) (Table 1).

The study intervention included all pregnant women living in the targeted areas, while the baseline and endline surveys included women aged 18–49 years in selected households who had given birth in the 9 months prior to each survey. The catchment area of these HFs included 113 villages, all of which were included in the cross-sectional survey, with sample size per village selected proportional to village size. At the village level, the Expanded Programme on Immunization (EPI) sampling method was used to select households where women would be interviewed [16]. At the center of the village, a direction was chosen by spinning a bottle on the ground. Households along the indicated path were visited and the questionnaire was administered to consenting, eligible women until the desired sample size was achieved.

A sample size of 360 women (30 women per health facility catchment area) who had delivered in the last 9 months was required for each survey to achieve 80% power to detect a difference between the group proportions of approximately 19.6%, from a baseline proportion of 44.7% (average proportion in the three districts) to an endline proportion of 64.3% in the intervention arm, as estimated using PASS V14 (NCSS, LLC, Kaysville, UT) to assess sample size for cluster randomized trial, at a significance level of 0.05 and assuming an intra-cluster correlation (ICC) of 0.03.

The MOH ensured the availability of SP tablets/treatment for IPTp in both control and intervention HFs throughout the study, including for distribution by trained CHWs. Each pregnant woman seen at ANC in the intervention facility received information on the community IPTp study and had the choice to participate or not. Pregnant women who wanted to receive SP had their identities recorded on a summary sheet and in CHW registers to facilitate their follow-up for the intervention. Women were asked to follow-up with the CHW monthly; those who did not present for the scheduled visit with the CHW were followed at home. The first dose of IPTp-SP was always given as directly observed therapy (DOT) at the HF to ensure that a health provider confirmed the woman’s gestational age and initiated IPTp after the end of the first trimester. CHWs also provided IPTp-SP as directly observed therapy, as part of their monthly household visits starting with the second dose, provided that SP had not been administered in the preceding 4 weeks. This was then recorded on the woman’s ANC card, which had space for recording up to five doses. CHWs also encouraged women to attend ANC, collected data on any adverse events, and referred women to the facility in the case of adverse events. CHW data were collected monthly. CHWs were supervised monthly by HF workers; monthly meetings were facilitated by the study.

Activities were implemented using the cascade training approach. The first training session included central-level stakeholders (National Malaria Control Programme, Directorate of Family Health, Directorate for Health Promotion and Education, Directorate of Sectorial Statistics) and the field coordinator from each district. The main objective of this training was to orient stakeholders on the organization and use of the various intervention tools. The stakeholder then transferred the information to the HF health workers, who in turn trained the CHWs (33 existing male and 25 existing female CHWs plus an additional 33 female volunteers).

For each survey, ten people were trained over 2 days in each district, and eight were chosen as field data collectors. The trainings focused on mastery of data collection tools, ethical rules, and ensuring data quality. CHWs and HF workers were trained on potential adverse events related to SP administration and instructed to report any adverse events to the study coordinator.

Routine service data (ANC visits, IPTp doses delivered by CHWs and at the HF) were summarized monthly to monitor the intervention. The primary outcome was the change in IPTp3 coverage over time assessed using a difference in difference analysis comparing the baseline and endline cross sectional survey data. In addition, a descriptive analysis of the socio-demographic characteristics of women included in the baseline and endline surveys, stratified by intervention and control areas, was performed. The REG procedure in SAS was used to assess for differences in medians. Difference in differences analyses were conducted on the survey data to assess the change over time in additional secondary outcomes, including any ANC visits (ANC1), four or more ANC visits (ANC4), IPTp1, 2, 4, and 5 coverage. Both primary and secondary outcomes were based on what was recorded on the ANC card, up to a maximum of five IPTp doses. The analyses used logistic regression with generalized estimating equations with the GENMOD procedure in SAS Version 9.4 (SAS Institute Inc., Cary, NC) for binary outcomes (ANC1, ANC4, IPTp3, IPTp4) with difference in differences calculated using a binary model with an identity link function, and Poisson regression for continuous outcomes (number of IPTp doses, number of ANC visits, timing of initiation of ANC). This was done by incorporating an intervention group variable, a variable indicating pre- or post-intervention, and an intervention-time interaction term. The ICC was calculated using the same model with the “corrw” option in the repeated statement. In order to obtain 95% confidence intervals (95% CI) for the difference in differences, a linear model was run using the GENMOD procedure, including all the same terms as in the logistic model. All analyses accounted for clustering at the HF level. Maternal age and gravidity were included a priori in the model; district and maternal education were explored as covariates but dropped due to lack of significance. A p-value of less than 0.05 was considered statistically significant.

The protocol was reviewed and approved by the John’s Hopkins School of Public Health (JHSPH) and Burkina Faso Institutional Review Boards (IRB); the Centers for Disease Control and Prevention Human Subjects Office determined that CDC staff were not engaged in human subjects research. Verbal informed consent was obtained from each woman in the local language prior to data collection.

Overall, 9834 doses were delivered to 2266 women living in the intervention HF catchment areas, with 50% of these given by CHWs, while 6683 doses were delivered to 2030 women living in the control HF catchment areas. There was improvement in the number of doses of IPTp delivered (Fig. 2), as well as in retention in care, measured by the number of women who completed four or more ANC visits out of the number of women who attended the first ANC visit (Fig. 3). During the period of the study, no adverse events related to SP administration were reported.

A total of 374 women were interviewed in the baseline survey and 360 in the endline survey. Socio-demographic characteristics of women in the control and intervention groups were similar both within and between surveys (Table 2). On average, a quarter of surveyed women had had one pregnancy, one-third had two to three pregnancies, and the rest had four or more.

At baseline, women received a median of 2.1 doses (range 0–5 doses) across both arms according to documentation on the ANC card, and a median of 2.5 doses (range 0–6) according to self-report, with no significant differences across arms. At endline, control group women received a median of 1.8 doses (range 0–5) according to the card and 2.8 doses (range 0–7) per self-report and women in the intervention group received 2.8 doses (range 0–5) per ANC card and 3.4 (range 0–7) per self-report (p-value < 0.0001 for difference in arms by both card and self-report).

ANC4, IPTp3, and IPTp 4 coverage were similar at baseline, with ANC4 coverage of 61.8% in the intervention and 62.2% in the control (p-value = 0.96), IPTp3 coverage of 50.5% in the intervention and 54.3% in the control (p-value = 0.74), and IPTp4 coverage of 21.5% in the intervention and 16.0% in the control (p-value = 0.59) (Table 3).

There was a 15.4 percentage point increase in the proportion of pregnant women who had attended at least four ANC visits in the intervention area, and a 2.8 percentage point increase in ANC4 attendance among women in the control group, for a DiD of 12.6 percentage points (95% CI − 7.0, 3.2). By the endline, administration of IPTp was universally higher in the intervention than control. There was an overall 17.6 percentage point increase in IPTp3 at endline in the intervention group, after accounting for baseline coverage (DiD 95% CI − 16.3, 51.5; p-value = 0.31). There was a 25.2 percentage-point increase in IPTp4 in the intervention area (p-value 0.02 for comparison from baseline to endline), and only a 5.1 percentage-point increase in the control (p-value 0.60 for comparison from baseline to endline), with a difference in differences of 20.0% (95% CI − 7.2, 47.3; p-value = 0.15). The greatest improvement was seen in IPTp5, which increased from 4.3% at baseline to 27.8% at endline in the intervention arm (DiD = 23.9 percentage points, 95% CI 9.4, 38.4; p = 0.001) (Table 3). Results did not change substantially when adjusted for maternal age and gravidity (Table 4). The calculated ICC was 0.095 for IPTp3 and 0.089 for IPTp4. Gravidity was associated with receipt of IPTp3, but not IPTp4 nor ANC attendance, while age greater than 20 was associated with ANC attendance (both ANC1 and ANC4), but not receipt of IPTp (Table 5).

At baseline, women reported first presenting to ANC at a mean of 3.1 and 2.8 months gestational age, in control and intervention arms, respectively (p-value = 0.07); while at endline, mean gestational age at first presentation to ANC was reported to be 3.0 and 2.6 months (p-value 0.03) in control and intervention arms, controlling for age and gravidity. Among the 90.1% of women who stated that they told someone about their pregnancy, the vast majority (91.7%) disclosed that they were pregnant prior to feeling the baby moving/showing, with no significant differences between baseline/endline or control and intervention groups. Two-thirds (65.9%) reported having first told their husbands/the father of the baby, 12.1% first told their mother in law, 8.2% first told a HF health worker, and 4.9% first told their mother; only 1.8% first reported their pregnancy to a CHW.

Adjusting for age and gravidity, at baseline 34.4% and 56.7% pregnant women in the control and intervention arms, respectively, reported having spoken to a CHW about their pregnancy; this increased to 54.0% and 77.6% in the control and intervention arms, respectively, at endline. The increase was statistically significant in both arms (p-value < 0.0001 in control and p-value 0.0028 in the intervention arm; p-value 0.0002 for the difference between arms at endline). Similar proportions reported discussing prevention of MIP with a CHW (26.7% and 57.9% in the control and intervention arms, respectively, at baseline, and 51.3% and 74.3% in the control and intervention arms, respectively, at endline). The increase was statistically significantly different in the control (p-value 0.007) but not the intervention arm (p-value 0.13), however, given the difference at baseline there was still a statistically significant difference between arms at endline (p-value 0.01).