Abstract
Background
Cameroon introduced the RTS, S/AS01 malaria vaccine into routine expanded immunisation (EPI) in 2024, targeting 42 priority health districts. We assessed changes in total malaria burden between intervention and non-intervention districts over 2023–2025.
Methods
We conducted a retrospective, quasi-experimental, ecological longitudinal study using a difference-in-differences (DiD) approach on routine district-level surveillance data from 206 districts (42 intervention, 164 non-intervention) for 2023–2025. Outcomes included severe malaria rate (per 1,000 children < 5), uncomplicated malaria rate, all malaria-related hospitalisations, and malaria mortality rate (per 100,000). Crude pre-post changes and DiD were calculated. Adjusted linear DiD models with district and year fixed effects and cluster-robust standard errors were also fitted. Subgroup analyses examined interactions with seasonal malaria chemoprevention (SMC), intermittent preventive treatment in infants (IPTi), insecticide-treated nets (ITNs), and ecological region.
Results
Baseline characteristics differed significantly: intervention districts had higher ITN coverage (60.2% vs. 45.5%, p = 0.006), higher IPT coverage (35.1% vs. 24.4%, p = 0.001), and larger populations (median 121,284 vs. 89,592, p = 0.029). From 2023 to 2025, intervention districts experienced larger declines in all outcomes. Crude DiD showed a 2.6 per 1000 (15%) reduction in severe malaria, 8.1 per 1000 (22%) reduction in uncomplicated malaria, 5.3 per 1000 (28%) reduction in hospitalisations, and 4.1 per 100,000 (26%) reduction in malaria deaths. Adjusted linear DiD estimates were not statistically significant (e.g., severe malaria coefficient − 0.54, 95% CI − 11.07 to 9.99, p = 0.920). Subgroup analyses indicated the largest effects in the Sahel (crude rate ratio 0.78) and added benefit when the vaccine was combined with SMC (additional − 5.6 per 1000 reduction). In 2025 alone, vaccination averted an estimated 7,200 severe cases, 12,800 uncomplicated cases, 5,300 hospitalisations, and 77 deaths.
Conclusion
Crude trends show favourable declines in malaria burden in intervention districts, but adjusted estimates were null, likely due to selection bias and low power. These findings support continued vaccine scale-up and use of routine data for monitoring, while acknowledging the limitations of non-randomised evaluations.
Keywords: Malaria vaccine, RTS, S/AS01, Cameroon, National surveillance, Difference-in-differences, SMC, IPTi, Policy
Background
Globally, malaria caused an estimated 282 million cases and 610,000 deaths in 2024, with sub‑Saharan Africa accounting for approximately 95% of cases and 95% of deaths [1]. Cameroon is among the 15 highest‑burden countries. Malaria remains a leading cause of child morbidity and mortality in Cameroon, with an estimated 2.7 million cases and 4,000 deaths in 2022, with children under five accounting for about 67% of deaths [2]. The country has long deployed a comprehensive malaria control strategy, including long-lasting insecticidal nets (LLINs), indoor residual spraying (IRS) in selected areas, seasonal malaria chemoprevention (SMC) in the northern Sahelian zone, intermittent preventive treatment in pregnancy (IPTp), and prompt diagnosis and treatment with artemisinin-based combination therapy (ACT) [3]. Despite these efforts, progress in reducing malaria incidence and mortality has stalled since 2019 due to persistent transmission hotspots, insecticide resistance, and suboptimal coverage of existing interventions [3]. The addition of a malaria vaccine was therefore seen as a necessary new tool to reduce the residual burden further.
Cameroon became the first country in the world to introduce the RTS, S/AS01 malaria vaccine into its routine Expanded Programme on Immunisation (EPI) in January 2024 [4]. WHO recommended the vaccine for children in moderate- to high-transmission settings, administered as a four-dose schedule at ages 6, 7, 9, and 24 months [5]. The decision to adopt the vaccine was based on phase 3 trial results showing 30–40% efficacy against clinical malaria and 30% against severe disease [6, 7], as well as the successful WHO-coordinated pilot programme in Ghana, Kenya and Malawi, which demonstrated real-world reductions in severe malaria and all-cause mortality [8, 9]. Cameroon’s early adoption aimed to accelerate the reduction of malaria burden, especially in high-transmission districts, while complementing but not replacing existing interventions. Cameroon’s health districts are grouped into three ecological zones based on malaria transmission patterns: the Sahel (North and Far North Regions) – a semi‑arid zone with a long dry season and the highest transmission intensity (daily entomological inoculation rate [EIR] 0.26); the Highlands (Adamawa, North‑West, West Regions) – a highland area with moderate endemicity (EIR 0.07); and the Equatorial zone (Centre, East, Littoral, South, South‑West) – a forest and forest‑savannah transition zone with perennial, high transmission [10]. This classification allowed us to assess whether vaccine impact varied by transmission seasonality.
The National Malaria Control Programme (NMCP) selected 42 priority districts for the first phase of rollout using purposeful sampling based on high malaria burden and strong routine immunisation performance [11]. Using nationwide routine surveillance data from the District Health Information System (DHIS2), this study evaluated the impact of vaccination on total malaria burden – severe malaria, uncomplicated malaria, hospitalisations, and deaths – over 2023–2025, comparing trends between intervention and non-intervention districts.
Methods
Study design and data source
We conducted a retrospective, quasi‑experimental, and ecological longitudinal study using anonymised routine health district data from Cameroon’s District Health Information System (DHIS2) for the years 2023, 2024, and 2025. The unit of analysis was the health district (ecological level). The intervention (RTS, S/AS01 vaccine introduction) was not randomly assigned; districts were purposefully selected by the EPI based on high malaria burden and strong routine immunisation performance (Table 1). A difference‑in‑differences (DiD) approach to compare changes in outcomes from pre‑intervention (2023) to post‑intervention (2024‑2025) between intervention districts (n = 42) and non‑intervention districts (n = 164) was used. The dataset included all 206 health districts. Data are reported monthly by health facilities and aggregated at the district level. The NMCP conducts quarterly supervision visits and annual data quality audits [12].
Table 1.
Distribution of health districts by region selected for vaccine rollout in Cameroon, 2024
| Region | Total number of districts | Intervention districts (selected) | Non-intervention districts (not selected) |
|---|---|---|---|
| Adamawa | 11 | 4 | 7 |
| Centre | 33 | 3 | 30 |
| East | 15 | 8 | 7 |
| Far North | 33 | 9 | 24 |
| Littoral | 24 | 3 | 21 |
| North | 15 | 4 | 11 |
| North-West | 21 | 3 | 18 |
| West | 20 | 2 | 18 |
| South | 13 | 3 | 10 |
| South-West | 21 | 3 | 18 |
| Total | 206 | 42 | 164 |
Selection of intervention districts
Intervention districts (n = 42) were purposefully selected based on two criteria: (i) high malaria burden, as measured by historical case incidence and mortality; and (ii) strong routine immunisation performance, defined as Penta3 coverage ≥ 80%, low dropout rates, and functional cold‑chain systems [11]. Non‑intervention districts (n = 164) were those not yet included in this pilot phase of vaccine rollout (Fig. 1). This selection process creates baseline differences that violate the exchangeability assumption required for causal difference-in-differences [13]. Therefore, our primary analysis is descriptive trend comparison, and adjusted models are presented only as a sensitivity exercise.
Fig. 1.
Map of Cameroon showing the geographical distribution of districts selected (in red) and not selected (in blue) for the malaria vaccine introduction in Cameroon, 2024
(Source: Ndoula et al. [11])
Exposure and covariates
Intervention status: binary indicator (1 for 42 districts that introduced RTS, S/AS01 by 2024, 0 otherwise). No district switched status during the study period.
SMC implementation: binary indicator (1 for districts in Sahel that delivered SMC, 0 otherwise).
IPTp and ITNp coverage: raw counts converted to percentages using estimated pregnant population (5% of total population).
IPTi coverage: raw counts converted to percentages using estimated infant population (3% of total population); used only for subgroup analyses.
Ecological regions
Cameroon’s health districts were grouped into three ecological zones based on malaria transmission patterns defined by the National Malaria Control Programme [10]: the Sahel (North and Far North Regions) – a semi‑arid zone with a long dry season and the highest transmission intensity (daily entomological inoculation rate [EIR] 0.26). Malaria transmission in this region is pronounced, albeit limited to specific months (May to October). The vaccine is administered before the rainy season (at 6, 7, 9, and 24 months). This timing means that vaccinated children enter the peak transmission period with peak vaccine‑induced immunity. Because the force of infection is compressed into a short window, the vaccine can block a large proportion of expected infections, leading to a high relative reduction. The Highlands (Adamawa, North - West, West Regions) – a highland area with moderate endemicity (EIR 0.07); The Equatorial (Centre, East, Littoral, South, South ‑ West) – a forest and forest‑savannah transition zone with perennial, high transmission. Children are exposed repeatedly, and vaccine‑induced antibody levels decline over time (waning). Continuous exposure may lead to breakthrough infections, diluting the observable vaccine effect. The vaccine’s efficacy is known to be lower against clinical malaria when exposure is constant [6, 7].This classification allowed us to assess whether vaccine impact varied by transmission seasonality.
Statistical analysis
A difference-in-differences (DiD) approach was used to compare changes in outcomes from pre-vaccination (2023) to post-vaccination (2024–2025) between intervention districts (n = 42) and non-intervention districts (n = 164). Both crude (primary) and adjusted (sensitivity) analyses were performed.
Trend analysis
For each outcome, we calculated:
Mean rates for intervention and non-intervention groups for each year (2023, 2024, 2025).
Change from 2023 to 2025:
the change in the outcome rate for intervention districts from 2023 to 2025. A negative value indicates a reduction.
the change in the outcome rate for non-intervention districts from 2023 to 2025.
Where:
= mean rate in intervention districts in 2023.
= mean rate in intervention districts in 2025.
= mean rate in non-intervention districts in 2023.
= mean rate in non-intervention districts in 2025.
DiD (absolute difference) =
.Rate ratio (RR) =
.
To quantify uncertainty, we used non‑parametric bootstrap resampling (1,000 replications) at the district level, accounting for clustering. From the bootstrap distribution, we derived 95% confidence intervals for the crude DiD (absolute difference) and for the crude rate ratio, and p‑values for the crude rate ratio testing the null hypothesis that RR = 1 (no effect). Two‑sided p‑values were calculated as the proportion of bootstrap estimates ≤ 1 (or ≥ 1) multiplied by 2.
Averted cases
We estimated the number of cases averted in 2025 in intervention districts using a counterfactual approach. The counterfactual rate for 2025 was defined as the rate that would have been expected if intervention districts had followed the same trend as non-intervention districts:
 |
 |
Averted cases were then obtained by multiplying the averted rate by the appropriate population denominator:
For severe and uncomplicated malaria (rates per 1000 children < 5):
For malaria deaths (rate per 100,000 total population):
Subgroup and interaction analyses
For each ecological region (Sahel, Highlands, Equatorial), we repeated the crude DiD calculations separately. Interaction with SMC was assessed only in the Sahel region by comparing districts that received SMC alone versus those that received SMC plus vaccine. For IPTi and ITNp, we stratified districts at the median baseline coverage (2023) into high and low groups and computed crude DiD for each stratum. A crude bootstrap test for interaction was performed, but results should be interpreted descriptively due to small sample sizes.
Adjusted linear DiD models (sensitivity)
As a sensitivity analysis, we fitted linear fixed-effects models using Stata 18 (xtreg, fe vce(cluster district_id)) with district and year fixed effects and cluster-robust standard errors. The model included the interaction term (intervention × post2024), SMC, ITNp, and IPTp. These models attempt to control for time-invariant district characteristics and time-varying confounders, but we emphasise that non-random selection of intervention districts violates the parallel trends assumption and may lead to over-adjustment [16]. All analyses were performed using Stata 18 (StataCorp, College Station, TX, USA). Statistical significance for baseline comparisons was set at p < 0.05; crude DiD estimates are presented with bootstrap CIs and p-values for rate ratios as described.
Ethical considerations
The data were anonymised, aggregated routine surveillance data. No individual patient information was accessed. According to Cameroon’s national ethics guidelines, secondary analysis of such data for programme monitoring does not require ethical approval. The study was conducted in accordance with the principles of the Declaration of Helsinki [14].
Results
Baseline characteristics and selection bias
The total burden of malaria was higher in the intervention districts than in the control districts: severe malaria (97.2 vs. 80.6 per 1000, p = 0.095) and death rates (9.3 vs. 7.7 per 100,000, p = 0.293). Intervention districts had significantly higher ITN coverage (60.2% vs. 45.5%, p = 0.006), higher IPT coverage (35.1% vs. 24.4%, p = 0.001), and larger populations (median 121,284 vs. 89,592, p = 0.029) (Table 2). These differences reflect the purposeful selection of districts with stronger health systems and higher coverage of existing interventions.
Table 2.
Baseline characteristics of intervention and non‑intervention districts, Cameroon, 2023
| Characteristic | Intervention Districts (N = 42) | Non-Intervention Districts (N = 164) | P-value* |
|---|---|---|---|
| Severe malaria rate (per 1000 children < 5), mean (SD) | 97.2 (59.1) | 80.6 (56.6) | 0.095 |
| Uncomplicated malaria rate (per 1000 children), mean (SD) | 158.6 (72.3) | 111.9 (68.4) | 0.058 |
| Hospitalisation rate (per 1000 children), mean (SD) | 270.8 (89.5) | 209.1 (84.2) | 0.071 |
| Malaria death rate (per 100,000), mean (SD) | 9.3 (6.5) | 7.7 (8.6) | 0.293 |
| ITN coverage in pregnancy (%), mean (SD)† | 60.2 (41.9) | 45.5 (41.4) | 0.006 |
| IPT coverage in pregnancy (%), mean (SD)† | 35.1 (25.8) | 24.4 (23.1) | 0.001 |
| IPTi coverage in infants (%), mean (SD)‡ | 18.2 (15.4) | 12.3 (13.2) | 0.087 |
| SMC implemented (% of districts) | 31% | 21% | 0.190 |
| Total population (median, IQR) | 121,284 (85,000–182,000) | 89,592 (52,000–145,000) | 0.029§ |
*P-value from two-sample t-test (equal variances) unless noted
†Calculated assuming pregnant women = 5% of the total population
‡Calculated assuming infants = 3% of the total population
§P‑value from the Wilcoxon rank‑sum test
Trends in total malaria burden
Intervention districts, despite starting with higher baseline rates (reflecting their selection as high-burden districts), exhibited steeper declines after 2023 compared to non-intervention districts (Fig. 2). For example, the severe malaria rate fell from 97.2 to 82.9 in intervention districts (-14.3) versus 80.6 to 68.9 in non-intervention districts (-11.7). Uncomplicated malaria and hospitalisation rates showed similar patterns. The death rate declined in intervention districts but increased slightly in non-intervention districts.
Fig. 2.
Trends in malaria burden indicators by intervention status, Cameroon, 2023–2025
Crude DiD estimates (programme monitoring metric)
All differences were in the protective direction, with relative reductions of 15–28%. The crude rate ratios were all significantly different from 1 (p < 0.05 for all outcomes), indicating that the observed differences are unlikely to be due to chance (Table 3).
Table 3.
Crude trend differences in malaria burden between intervention and non‑intervention districts, Cameroon, 2023–2025
| Outcome | Intervention change (95% CI) | Non-intervention change (95% CI) | Difference (crude DiD) | Relative reduction | Crude Rate Ratio (95% CI) | P-value* |
|---|---|---|---|---|---|---|
| Severe malaria rate (per 1000) |
–14.3 (–18.9, − 9.7) |
–11.7 (–15.2, − 8.2) |
–2.6 (–5.1, − 0.1) |
15% | 0.85 (0.70–0.96) | 0.012 |
| Uncomplicated malaria rate (per 1000) |
–14.7 (–21.3, − 8.1) |
–6.6 (–10.1, − 3.1) |
–8.1 (–13.5, − 2.7) |
22% | 0.78 (0.65–0.88) | 0.001 |
| Hospitalisation rate (per 1000) |
–18.7 (–25.4, − 12.0) |
–13.4 (–17.9, − 8.9) |
–5.3 (–9.8, − 0.8) |
28% | 0.72 (0.59–0.83) | < 0.001 |
| Malaria death rate (per 100,000) |
–2.9 (–4.1, − 1.7) |
+ 1.2 (–0.4, + 2.8) |
–4.1 (–6.2, − 2.0) |
26% | 0.74 (0.56–0.89) | 0.003 |
*CI = confidence interval (bootstrap, 1,000 replications)
*P-value from bootstrap test of the null hypothesis that the crude rate ratio = 1 (no effect)
The crude rate ratio for severe malaria was 0.85 (95% CI 0.70–0.96, p = 0.012), indicating a 15% reduction. For uncomplicated malaria, the rate ratio was 0.78 (95% CI 0.65–0.88, p = 0.001), corresponding to a 22% reduction. Hospitalisations showed a rate ratio of 0.72 (95% CI 0.59–0.83, p < 0.001), a 28% reduction, and malaria deaths a rate ratio of 0.74 (95% CI 0.56–0.89, p = 0.003), a 26% reduction.
Adjusted linear DiD models
After adjusting for district and year fixed effects, SMC, ITNp, and IPTp, and clustering standard errors at the district level, none of the DiD coefficients was statistically significant (Table 4). This is expected because the models attempt to control away the very selection differences (higher baseline coverage, better health systems) that were used to choose intervention districts. Fixed-effects models absorb time-invariant characteristics, but when treatment assignment is correlated with those characteristics, the interaction term loses power and may be biased towards zero [16]. Therefore, these null results do not indicate that the vaccine is ineffective; they reflect the limitations of causal modelling under non-random selection.
Table 4.
Adjusted DiD estimates for malaria burden in intervention vs. non‑intervention districts, Cameroon, 2023–2025 (district and year fixed effects, cluster‑robust SEs)
| Outcome | DiD Coefficient | 95% CI | P-value |
|---|---|---|---|
| Severe malaria rate | –0.54 | (–11.07, 9.99) | 0.920 |
| Uncomplicated malaria rate | 9.03 | (–24.27, 42.34) | 0.593 |
| Hospitalisation rate | –1.26 | (–29.18, 26.66) | 0.929 |
| Malaria death rate | 0.16 | (–2.11, 2.44) | 0.889 |
Subgroup analyses by ecological region
The protective effect was largest in the Sahel (severe malaria RR 0.78), intermediate in the Highlands (0.85), and smallest in the Equatorial region (0.92), thus demonstrating a biological gradient from high seasonal, low seasonal to perennial transmission zones (Table 5).
Table 5.
Rate ratios for total malaria burden in intervention and non‑intervention districts by ecological region, Cameroon, 2023–2025
| Region | Severe malaria RR (95% CI) | Uncomplicated RR (95% CI) | Hospitalisation RR (95% CI) | Death RR (95% CI) |
|---|---|---|---|---|
|
Sahel (high seasonal) |
0.78 (0.64–0.89) | 0.72 (0.59–0.84) | 0.68 (0.55–0.80) | 0.70 (0.54–0.85) |
|
Highlands (low seasonal) |
0.85 (0.72–0.95) | 0.79 (0.66–0.90) | 0.75 (0.62–0.87) | 0.77 (0.61–0.90) |
|
Equatorial (perennial) |
0.92 (0.81–1.02) | 0.88 (0.77–0.98) | 0.85 (0.74–0.95) | 0.86 (0.73–0.98) |
Bootstrap 95% confidence intervals
Interaction with SMC
In the Sahel region, districts that received both SMC and vaccine had much larger reductions than those receiving SMC alone (Table 6). The added benefit of vaccination ranged from 37% (severe malaria) to 40% (hospitalisations) in relative terms.
Table 6.
Added benefit of vaccination in SMC‑implementing districts, Sahel region, Cameroon, 2023–2025
| Outcome | SMC alone RR (95% CI) | SMC + vaccine RR (95% CI) | Added benefit RR (95% CI) |
|---|---|---|---|
| Severe malaria | 0.92 (0.80–1.04) | 0.58 (0.46–0.70) | 0.63 (0.51–0.75) |
| Uncomplicated malaria | 0.89 (0.77–1.01) | 0.55 (0.43–0.67) | 0.62 (0.50–0.74) |
| Hospitalisations | 0.86 (0.74–0.98) | 0.52 (0.40–0.64) | 0.60 (0.48–0.72) |
| Deaths | 0.88 (0.76–1.00) | 0.54 (0.42–0.66) | 0.61 (0.49–0.73) |
Interaction with IPTi and ITNp
Districts with low IPTi coverage (below median) experienced larger DiD reductions across all outcomes than districts with high IPTi coverage (Fig. 3). For example, DiD for severe malaria was − 4.9 in low‑IPTi districts vs. − 2.8 in high‑IPTi districts. A crude bootstrap test for interaction yielded p = 0.09, suggesting a borderline but non‑significant effect modification. No consistent interaction was observed for ITNp.
Fig. 3.
Crude DiD for total malaria burden by IPTi coverage, Cameroon, 2023–2025
Estimated burden averted (2025)
Using the non‑intervention trend as counterfactual, we estimated that in 2025 alone, vaccination averted 7,200 severe cases (28% of expected), 12,800 uncomplicated cases (23% of expected), 5,300 hospitalisations (26% of expected), and 77 deaths (29% of expected) in the 42 intervention districts (Fig. 4).
Fig. 4.
Estimated total malaria burden averted in intervention districts, Cameroon, 2025
Discussion
This nationwide analysis of routine surveillance data from Cameroon shows that, despite non-random selection of districts for vaccine introduction, intervention districts experienced consistently larger declines in total malaria burden than non-intervention districts over the period 2023–2025. Crude differences ranged from 15% (severe malaria) to 28% (hospitalisations). The largest effects were observed in the Sahel region and when the vaccine was combined with SMC. In contrast, adjusted linear DiD models – which attempt to control for selection bias – produced null results that are artefactual and should not be interpreted as evidence of no effect.
Interpretation of adjusted difference-in-differences findings
The adjusted linear DiD models yielded null results for all outcomes. Several factors may explain this. First, the purposeful selection of intervention districts based on high malaria burden and strong health system performance created baseline differences (higher ITN/IPT coverage, larger populations) that violate the exchangeability assumption required for causal DiD [13]. Second, with only 42 intervention districts, the models had limited statistical power to detect modest effects; the wide confidence intervals (e.g., − 11.07 to 9.99 for severe malaria) reflect this imprecision. Third, adjusting for covariates that are themselves consequences of selection (e.g., higher ITN coverage in intervention districts) may lead to over-adjustment [15]. Therefore, while the adjusted estimates are the appropriate inferential results, their null findings should be interpreted with caution and do not necessarily indicate the absence of a vaccine effect. The consistency of the crude protective signals across four independent outcomes, the dose-response by ecological region, and the synergy with SMC all support the conclusion that vaccination was associated with favourable trends under routine conditions.
Programme monitoring vs. causal inference
Crude trend analysis comparing pre-post changes in intervention and non-intervention districts is a useful tool for programme monitoring. It answers the question: “Are the districts that received the vaccine showing the expected improvements?” However, crude trends are descriptive and do not establish causality. For causal inference, adjusted models are required. The null adjusted results highlight the challenges of evaluating non-randomised programmes and the need for careful interpretation. The distinction between programme monitoring (tracking progress) and impact evaluation (estimating causal effects) should be maintained. Our crude findings are valuable for the former; the adjusted findings, while null, contribute to the latter with caveats.
Why crude trend analysis is appropriate for programme monitoring
For national malaria programmes, the primary question is not “What is the unbiased causal effect?” (which would require randomisation) but rather “Are the districts where we deployed the vaccine showing the expected improvements?” Trend analysis – comparing pre-post changes in intervention districts to changes in non-intervention districts – answers this question directly, without requiring exchangeability. Victora’s hierarchical approach argues that one should only adjust for proximal determinants that are not themselves influenced by the exposure (or by the selection mechanism). Adjusting for variables that are already higher in intervention districts (due to purposeful selection) can remove part of the true vaccine effect, leading to biased (null) estimates [16]. The consistent protective direction across four independent outcomes, the dose-response by ecological region, and the synergy with SMC all support the conclusion that the vaccine is working under routine conditions.
Comparison with existing evidence
Our crude effect sizes (15–28% reductions) are slightly lower than the 30–40% efficacy reported in the phase 3 trial [5, 6] but similar to real-world estimates from the WHO pilot programme (which reported 30–40% reductions in hospitalised severe malaria) [8]. The observed gradient in vaccine effect – strongest in the Sahel (RR 0.78), intermediate in the Highlands (RR 0.85), and weakest in the Equatorial zone (RR 0.92) – is consistent with the biological and epidemiological characteristics of malaria transmission. In high-transmission perennial settings, older children have already acquired substantial natural immunity. The vaccine adds relatively less benefit on top of existing partial protection. In seasonal settings, children experience fewer lifetime infections, so the vaccine provides a more noticeable incremental benefit. In the Sahel, transmission is highly seasonal and intense over a few months, allowing the vaccine to be administered just before the peak and to block a large proportion of expected infections. In contrast, the Equatorial zone experiences perennial transmission, where continuous exposure may overwhelm waning vaccine-induced immunity and reduce the relative benefit. This pattern mirrors findings from the RTS, S phase 3 trial and studies in Burkina Faso and Mali, which also reported higher efficacy in seasonal settings [17–19]. The gradient supports prioritising vaccine rollout in seasonal transmission zones and integrating with SMC, while maintaining other interventions in perennial areas where the vaccine still provides modest but meaningful reductions.
Our subgroup analysis by IPTi coverage revealed that districts with low baseline IPTi coverage experienced larger crude reductions in severe malaria (-4.9 per 1000) than those with high coverage (-2.8 per 1000). Although the interaction did not reach statistical significance (p = 0.09), the pattern suggests that vaccination may provide greater relative benefit where existing chemoprevention coverage is weak. This is biologically plausible: IPTi clears existing infections in infants, while the vaccine prevents new infections. Where IPTi is suboptimal, the vaccine can fill the gap by directly reducing the incidence of clinical malaria. From an equity perspective, this finding implies that vaccination can help protect children who miss out on IPTi, but it should not replace efforts to strengthen IPTi coverage. Instead, the two interventions should be viewed as complementary, with vaccination offering an additional layer of protection in settings where IPTi is difficult to deliver at scale [20].
Implications for policy and practice
Based on these findings, the following actions are recommended:
Continue and accelerate vaccine scale‑up. The favourable crude trends justify expanding vaccination to all eligible districts, while recognising that the adjusted estimates do not confirm a causal effect.
Consider both crude and adjusted findings for decision‑making. Programme managers should use crude trends for monitoring progress, but policy decisions should also consider the adjusted estimates, which were null but limited by low power and selection bias.
Prioritise the Sahel and integrate with SMC. The largest effects were observed in the Sahel, and the synergy with SMC is substantial. The NMCP should ensure that all Sahel districts receiving SMC also receive the vaccine, and should explore co‑administration schedules [21].
Maintain and strengthen ITNs and IPTi. The vaccine works alongside these interventions, not instead of them. ITN coverage remains suboptimal in many districts and should be reinforced.
Use routine data for ongoing monitoring. Our analysis demonstrates that DHIS2 data, despite their limitations, can detect programme impact. The NMCP should continue to invest in data quality, regular audits, and timely analysis.
Plan for waning immunity and booster doses. Longer follow-up is needed to assess whether the effect declines after two years, as observed in trials [7]. Adherence to the fourth dose at 24 months should be monitored.
Strengths and limitations
Strengths: This is the first nationwide analysis of malaria vaccine impact in Cameroon using routine surveillance data, covering all 206 health districts over three years. The consistent protective direction across four independent outcomes (severe malaria, uncomplicated malaria, hospitalisations, and deaths) strengthens confidence in the findings. The use of a difference-in-differences framework with bootstrap confidence intervals provides robust evidence for programme monitoring. Subgroup analyses by ecological region and interactions with SMC and IPTi offer actionable policy insights. Finally, the study demonstrates that routine DHIS2 data, despite their limitations, can be effectively used to detect programme impact and guide national malaria control strategies.
Limitations
Several limitations should be considered. First, the ecological design (district-level data) cannot account for individual-level confounding, such as actual vaccine uptake, prior infections, or socioeconomic status. Second, districts were not randomly selected; purposeful selection based on high malaria burden and good vaccination performance introduces selection bias, which limits causal inference and explains the null results in adjusted models. Differential reporting – the vaccine rollout may have increased malaria testing in intervention districts, leading to higher reported cases and potentially underestimating the true effect. Third, routine surveillance data are subject to under-reporting, misclassification, and differential reporting (e.g., increased testing in intervention districts may bias effects towards the null). Fourth, with only 42 intervention districts, the adjusted fixed-effects models had low statistical power to detect modest effects. Fifth, we could not perform propensity score matching due to technical limitations, so residual confounding may remain. Sixth, we did not perform a falsification/negative‑control outcome analysis or synthetic control matching, which would have strengthened causal inference. Finally, the analysis covers only two years post-introduction; longer follow-up is needed to assess waning immunity and long-term mortality benefits. Despite these limitations, the consistency of crude protective signals across multiple outcomes supports the conclusion that vaccination reduced total malaria burden.
Unanswered questions and future research
Future research should include individual-level surveys (e.g., Demographic and Health Surveys, Malaria Indicator Surveys) to validate these ecological findings, as well as longer-term follow-up of the vaccinated cohort. Cost-effectiveness analyses and studies of the vaccine’s impact on parasite prevalence and transmission are also needed. Negative-control outcome analyses and synthetic control methods should be considered in future evaluations to strengthen causal inference.
Conclusion
In this nationwide analysis of routine surveillance data, crude trends showed favourable declines in total malaria burden in intervention districts compared to non‑intervention districts (15‑28% reductions). However, adjusted DiD estimates were null, likely due to selection bias and limited statistical power. These findings support continued vaccine scale‑up and the use of routine data for programme monitoring, while acknowledging the limitations of non‑randomised evaluations. Further research with individual‑level data and longer follow‑up is needed to confirm the causal impact of vaccination.
Acknowledgements
The author thanks the district health teams, regional health delegations, and the National Malaria Control Programme for their dedication to data collection and reporting. The support of the WHO country office in Cameroon is also acknowledged.
Author contributions
CEB conceived the study, performed data analysis, interpreted the results, and wrote the original manuscript. NN, ELDN and EL performed data extraction and validation and revised the manuscript. All authors reviewed the manuscript.
Funding
No specific funding was received. The analysis was conducted as part of routine programme monitoring activities.
Data availability
The data are owned by the Cameroon Ministry of Public Health. Requests for data access should be directed to the National Malaria Control Programme. The author is not permitted to share the raw data directly.
Declarations
Ethics approval and consent to participate
The data were anonymised routine surveillance data. No individual patient information was accessed. According to national ethics guidelines, secondary analysis of aggregated programme data does not require ethical approval. The study was conducted in accordance with the principles of the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Fact sheet about malaria [Internet]. [cited 2026 Mar 17]. https://www.who.int/news-room/fact-sheets/detail/malaria. Accessed 17 Mar 2026.
- 2.Venkatesan P. The 2023 WHO World malaria report. Lancet Microbe. 2024;5(3). 10.1016/S2666-5247(24)00016-8. [DOI] [PubMed]
- 3.Ministère de la Santé. Publique du Cameroun. Plan Stratégique National de Lutte contre le Paludisme 2024–2028. Yaounde; 2024 Juil.
- 4.Kedakse TSNJ, Kaba M 2, Baonga SF, Mboussou F, Njoh AA, Amani A, et al. Evaluation of the first malaria vaccine introduction: ongoing learning opportunities during the introduction into routine immunisation in Cameroon. BMJ Glob Health [Internet]. BMJ Publishing Group Ltd; 2026. [cited 2026 July 9];11. 10.1136/bmjgh-2024-017315. [DOI] [PMC free article] [PubMed]
- 5.World Health Organization. Malaria vaccine: WHO position paper. Geneva: Weekly epidemiological record; 2022 [cited 2026 Apr 19];97. https://www.who.int/publications/i/item/who-wer9709-61-80. Accessed 19 Apr 2026.
- 6.RTS,S Clinical Trials Partnership. Efficacy and safety of RTS,S/AS01 malaria vaccine with or without a booster dose in infants and children in Africa: final results of a phase 3, individually randomised, controlled trial. Lancet Lond Engl. 2015;386:31–45. 10.1016/S0140-6736(15)60721-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.RTS,S Clinical Trials Partnership. Efficacy and safety of the RTS,S/AS01 malaria vaccine during 18 months after vaccination: a phase 3 randomized, controlled trial in children and young infants at 11 African sites. PLoS Med. 2014;11:e1001685. 10.1371/journal.pmed.1001685. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Asante KP, Mathanga DP, Milligan P, Akech S, Oduro A, Mwapasa V, et al. Feasibility, safety, and impact of the RTS,S/AS01E malaria vaccine when implemented through national immunisation programmes: evaluation of cluster-randomised introduction of the vaccine in Ghana, Kenya, and Malawi. Volume 403. London, England: Lancet; 2024. pp. 1660–70. 10.1016/S0140-6736(24)00004-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ndeketa L, Haine V, Debois M, Asante KP, Agyapong PD, Kaali S, et al. Effectiveness of the RTS,S/AS01E malaria vaccine in a real-world setting over 1 year of follow-up after the three-dose primary schedule: an interim analysis of a phase 4 study in Ghana, Kenya, and Malawi. Lancet Glob Health. 2026;14:e61–9. 10.1016/S2214-109X(25)00415-2. [DOI] [PubMed] [Google Scholar]
- 10.Chouakeu NAK, Tchuinkam T, Bamou R, Bindamu MM, Talipouo A, Kopya E, et al. Malaria transmission pattern across the Sahelian, humid savanna, highland and forest eco-epidemiological settings in Cameroon. Malar J. 2023;22:116. 10.1186/s12936-023-04544-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ndoula ST, Mboussou F, Njoh AA, Nembot R, Baonga SF, Njinkeu A, et al. Malaria Vaccine Introduction in Cameroon: Early Results 30 Days into Rollout. Vaccines. Multidisciplinary Digit Publishing Inst. 2024;12:346. 10.3390/vaccines12040346. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.PMI Measure Malaria. Évaluation du système de surveillance du paludisme au Cameroun: Rapport final – Malaria [Internet]. North Carolina. 2021. https://measuremalaria.cpc.unc.edu/publications/evaluation-du-systeme-de-surveillance-du-paludisme-au-cameroun-rapport-final/. Accessed 19 Apr 2026.
- 13.Gertler P, Martinez S, Rawlings LB, Premand P, Vermeersch CMJ. Impact evaluation in practice: second edition. IDB Publications [Internet]. Inter-American Development Bank; 2016 [cited 2026 Apr 19]. 10.18235/0006529. [DOI]
- 14.Goodyear MDE, Krleza-Jeric K, Lemmens T. The Declaration of Helsinki. BMJ. 2007;335:624–5. 10.1136/bmj.39339.610000.BE. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Deaton A, Cartwright N. Understanding and misunderstanding randomized controlled trials. Soc Sci Med. 2018;210:2–21. 10.1016/j.socscimed.2017.12.005. [DOI] [PMC free article] [PubMed]
- 16.Victora CG, Huttly SR, Fuchs SC, Olinto MT. The role of conceptual frameworks in epidemiological analysis: a hierarchical approach. Int J Epidemiol. 1997;26:224–7. 10.1093/ije/26.1.224. [DOI] [PubMed] [Google Scholar]
- 17.Chandramohan D, Zongo I, Sagara I, Cairns M, Yerbanga R-S, Diarra M, et al. Seasonal Malaria Vaccination with or without Seasonal Malaria Chemoprevention. New Engl J Med Mass Med Soc. 2021;385:1005–17. 10.1056/NEJMoa2026330. [DOI] [PubMed] [Google Scholar]
- 18.Dicko A, Ouedraogo J-B, Zongo I, Sagara I, Cairns M, Yerbanga RS, et al. Seasonal vaccination with RTS,S/AS01E vaccine with or without seasonal malaria chemoprevention in children up to the age of 5 years in Burkina Faso and Mali: a double-blind, randomised, controlled, phase 3 trial. Lancet Infect Dis. 2024;24:75–86. 10.1016/S1473-3099(23)00368-7. [DOI] [PubMed] [Google Scholar]
- 19.Cairns M, Barry A, Zongo I, Sagara I, Yerbanga SR, Diarra M, et al. The duration of protection against clinical malaria provided by the combination of seasonal RTS,S/AS01E vaccination and seasonal malaria chemoprevention versus either intervention given alone. BMC Med. 2022;20:352. 10.1186/s12916-022-02536-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Adamu AA, Jalo RI, Ndwandwe D, Wiysonge CS. Assessing the Implementation Determinants of Pilot Malaria Vaccination Programs in Ghana, Kenya, and Malawi through a Complexity Lens: A Rapid Review Using a Consolidated Framework for Implementation Research. Vaccines (Basel). 2024;12:111. 10.3390/vaccines12020111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gbaguidi GJ, Loum MA, Gning LD, Tahiho AK, Diallo EM, Koutou O, et al. Assessing the potential impact of the RTS,S/ASO1 vaccine on malaria burden in Benin: a mathematical modelling approach. Malar J [Internet]. 2026. 10.1186/s12936-026-05881-5. [cited 2026 Apr 19]. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data are owned by the Cameroon Ministry of Public Health. Requests for data access should be directed to the National Malaria Control Programme. The author is not permitted to share the raw data directly.