The unprecedented deployment of malaria interventions since 2000 led to a considerable decrease in malaria morbidity and mortality. However, progress in malaria control has stagnated in recent years. WHO estimated that 241 million cases of malaria occurred in 2020, resulting in 627 000 deaths, of which 77% were in children younger than 5 years and approximately 95% occurred in the WHO African region.¹ The efficacy of existing preventive interventions and life-saving therapies is threatened by the development of insecticide resistance by the vector and drug resistance by the parasite, and thus new tools are needed.

RTS,S/AS01_E (hereafter referred to as RTS,S; GSK, Wavre, Belgium) is the only vaccine currently recommended against malaria. A phase 3 clinical trial was conducted at 11 sites across seven sub-Saharan Africa countries between 2009 and 2014. Vaccine efficacy against all episodes of clinical malaria in children aged 5–17 months when administered according to a 0, 1, and 2 month primary schedule (M012) was 55·1% (95% CI 50·5 to 59·3) over 12 months of follow-up.² Over 4 years of follow-up, vaccine efficacy against clinical malaria was 28·3% (95% CI 23·3 to 32·9) and against severe malaria was 1·1% (–23·0 to 20·5); the addition of a fourth dose at month 20 (M012–20) increased vaccine efficacy against clinical malaria to 36·3% (31·8 to 40·5) and against severe malaria to 32·2% (13·7 to 46·9).³ These results supported pilot implementation of RTS,S through the Expanded Programmes on Immunisation in Ghana, Kenya, and Malawi.⁴ On Oct 6, 2021, WHO recommended widespread use of RTS,S in children in sub-Saharan Africa and areas with moderate-to-high Plasmodium falciparum transmission.⁵

Concomitantly, efforts continue to improve vaccine efficacy, durability of protection, and availability of RTS,S. Several controlled human malaria infection (CHMI) studies in malaria-naive adults have suggested that a regimen containing a fractional dose of different RTS,S formulations confers high protection against P falciparum infection.^6–9 One CHMI study found a vaccine efficacy of 86·7% (95% CI 66·8–94·6) with a delayed third fractional vaccine dose (one fifth of the full dose) at month 7 (M017) compared with 62·5% (29·4–80·1) with the standard M012 full-dose regimen, at month 8 after CHMI.⁷ These findings could result in substantial public health impact if corroborated in children in the setting of natural exposure, because a regimen with a fractional third or fourth dose might result in more children being vaccinated with RTS,S than when using a standard full-dose regimen. This has the potential to improve population-level vaccine availability, and thus reduce malaria morbidity and mortality.

We report results up to 20 months of follow-up of an ongoing phase 2b trial that aims to establish proof-of-concept for the use of fractional-dose regimens of RTS,S in children aged 5–17 months at first vaccination, under conditions of natural exposure. Therefore, we compared a fractional-dose regimen to the currently recommended standard full-dose schedule (M012). We also explored the effect of a delayed third dose in a fractional-dose regimen (M017), and the effect of an early full or fractional fourth dose (at month 14) on vaccine efficacy, over 1 year of follow-up after dose three and up to month 20. Immunogenicity and safety of all doses were also assessed up to month 21.

Between Sept 28, 2017, and Sept 25, 2018, 2157 children were enrolled, of whom 1609 were randomly assigned to a treatment group (322 to R012–20, 322 to R012–14, 322 to Fx012–14, 322 to Fx017–20, and 321 to the rabies vaccine control group). 1500 children received at least one study vaccine dose and the per-protocol set comprised 1332 children (figure 1). Baseline characteristics were similar across groups and countries and between the exposed set and the per-protocol set (table 1).

During the 12 months after dose three, the incidence of the first or only episode of clinical malaria meeting the primary case definition was 0·36 per person-year in the pooled R012–14 plus R012–20 groups, 0·45 per person-year in the Fx012–14 group, and 0·61 per person-year in the control group. The incremental vaccine efficacy in the Fx012–14 group versus the pooled R012–14 plus R12–20 groups was −21% (95% CI −57 to 7; p=0·15; figure 2). Superiority of the Fx012 regimen over the full-dose R012 schedule was not shown.

Over the 12 months after dose three, vaccine efficacy against first or only episode of clinical malaria meeting the primary case definition was 47% (95% CI 31–59) in the pooled R012–14 plus R012–20 groups, 35% (13–51) in the Fx012–14 group, and 54% (38–66) in the Fx017–20 group. Vaccine efficacy estimates were similar for each group irrespective of whether they were assessed for first or only episodes or all episodes and meeting the primary or secondary case definitions (figure 2; appendix pp 11–13).

Vaccine efficacy against all episodes of clinical malaria (secondary case definition) from day 0 to month 20 after four vaccinations in the R012–14 group and the Fx012–14 group, and after three vaccinations in the R012–20 group and the Fx017–20 group is shown in figure 2. The vaccine efficacy for the Fx017–20 regimen tended to be lower than that for M012 schedules at shorter follow-up periods (7 months after dose two or three), but the difference decreased over time up to month 20 (appendix pp 11–13). An estimated 213 (7%) of the 2843 malaria cases (all episodes, secondary case definition) were co-infected with Plasmodium malarie (143 [5%]) or Plasmodium ovale (74 [3%]) by microscopy.

Over the 12 months after dose three, incremental vaccine efficacy of a third fractional dose using the secondary case definition (all episodes) was −13% (95% CI −46 to 12) in the Fx012–14 group compared with the pooled R012–14 plus R012–20 groups (figure 2). Incremental vaccine efficacy against clinical malaria was 10% (–28 to 36) in the Fx017–20 group compared with the Fx012–14 group over 12 months after dose three (appendix p 14).

Parasite density was similar between study groups (appendix p 15). Estimates of vaccine efficacy against first or only episode of incident P falciparum infection up to month 20 in all RTS,S groups are shown in appendix p 16. Estimates of vaccine efficacy against all episodes of prevalent P falciparum infections are shown in appendix p 17.

Up to month 20, 1490 cases of clinical malaria (secondary case definition) per 1000 vaccinated children were averted in the Fx012–14 group and 1288 cases per 1000 vaccinated children were averted in the Fx017–20 group. The number of averted cases was different between the groups receiving the full-dose regimens (figure 3; appendix p 18). However, in a post-hoc analysis, the incidence of clinical malaria by 3-month periods was shown not to differ significantly (overlapping 95% CIs) between the R012–20 and R012–14 groups up to month 12 (appendix p 19).

The underlying incidence of malaria differed substantially between countries, as shown by the event rates in the control groups, with Kenya consistently having more events than Ghana. The difference was more marked for all episodes (secondary case definition) than for first or only episode (primary case definition) of clinical malaria. Vaccine efficacy and incremental vaccine efficacy estimates varied between Ghana and Kenya for all endpoints (figure 4, appendix pp 11–13). Point estimates of vaccine efficacy against first or only episode of incident P falciparum infections tended to be higher in Ghana (where incidence was lower); the same was observed for vaccine efficacy against all episodes of prevalent infections (appendix pp 16–17).

Up to month 20, the number of clinical malaria cases averted (figure 3) in RTS,S groups was higher in Kenya where the incidence of clinical malaria cases in the control group was approximately 6 times higher compared to Ghana (figure 4).

The distribution of anti-circumsporozoite protein antibody concentrations after vaccination is shown in appendix p 5. 1 month after dose three, anti-circumsporozoite protein antibody GMCs tended to be higher in the Fx012–14 than in the Fx017–20 group but similar GMCs were observed between the groups 1 month after dose four. Overall, the fourth dose did not increase anticircumsporozoite responses to levels observed after dose three (appendix pp 6, 20). There was no difference in anti-circumsporozoite protein antibody avidity between the fractional-dose and the full-dose regimens (appendix p 6). Anti-HBs antibody responses increased after each RTS,S administration (appendix pp 6, 20).

A summary of reported adverse events is shown in table 2. The incidence of solicited adverse events after doses three and four in the reactogenicity subset was low and similar between groups. Fever was the most frequent general adverse event, reported in both full-dose and fractional-dose groups: in five (12%) of 41 children in the R012–20 group, 11 (25%) of 44 in the R012–14 group, six (14%) of 42 in the Fx012–14 group, and two (5%) of 43 in the Fx017–20 group after dose four (table 2, appendix p 21). No clinically significant changes in haematology or biochemistry parameters were observed.

Unsolicited adverse events occurring within 30 days from any vaccination up to month 21 are shown in table 2 and appendix p 22, with upper respiratory tract infection (in 161 [54%] of 298 children in the R012–20 group, 175 [60%] of 294 in the R012–14 group, 184 [61%] of 304 in the Fx012–14 group, 181 [58%] of 311 in the Fx017–20 group, and 164 [56%] of 293 in the control group) and gastroenteritis (in 57 [19%] children in the R012–20 group, 81 [28%] in the R012–14 group, 72 [24%] in the Fx012–14 group, 67 [22%] in the Fx017–20 group, and 72 [25%] in the control group) being the most frequently reported.

Serious adverse events up to month 21 were reported in 48 (16%) children in the R012–20 group, 45 (15%) in the R012–14 group, 47 (15%) in the Fx012–14 group, 62 (20%) in the Fx017–20 group, and 71 (24%) in the control group (table 2, appendix p 22). Four deaths, unrelated to vaccination, were reported, caused by drowning in one boy (R012–20 group), gastroenteritis in two girls (R012–14 and Fx017–20 groups), and a wall falling on a boy (Fx017–20 group).

Malaria cases up to month 21 were reported in 89 (30%) children in the R012–20 group, 86 (29%) in the R012–14 group, 84 (28%) in the Fx012–14 group, 100 (32%) in the Fx017–20 group, and 101 (35%) in the control group (appendix p 22). Severe malaria was reported in 13 (4%) of 298 children in the R012–20 group, 14 (5%) of 294 in the R012–14 group, 15 (5%) of 304 in the Fx012–14 group, 20 (6%) of 311 in the Fx017–20 group, and 31 (11%) of 293 in the control group. Cerebral malaria was reported in one (<1%) child in the control group.

Cause-confirmed viral meningitis was detected in six children: one (<1%) each in the R012–20 and Fx012–14 groups and two (1%) each in the control and Fx017–20 groups. At least one seizure episode within 30 days after any vaccination occurred in 12 children (four [1%] in each of the R012–20, Fx017–20, and control groups). Convulsive seizures levels 1–2 (according to the Brighton Collaboration Working Group case definition for generalised convulsive seizures)¹⁸ within 7 days of any vaccination were reported for four (1%) children in the R012–20 group, four (1%) in the Fx017–20 group, and two (<1%) in the control group. A potential immune-mediated disease (blister; verbatim term bullous skin disease) occurring 245 days after dose three (R012–14 group) was not considered related to vaccination.

A post-hoc analysis showed that the prevalence of P falciparum infections varied greatly by calendar month from enrolment to study month 21, with no clear seasonal pattern observed (figure 5). Due to the difference in the start of enrolment, the overall prevalence reflects mainly data from Ghana over the first study months and from Kenya in the last study months. The greater variability observed in the earliest and latest calendar months was due to a lower number of children at risk or the total number of children included in the analyses. The prevalence of P falciparum infections at cross-sectional visits is shown in appendix p 7.

To our knowledge, this is the first study to assess whether regimens including fractional doses increase the protective efficacy and impact of RTS,S against malaria in countries with moderate-to-high malaria endemicity. Our findings show the use of a fractional instead of a full third RTS,S dose does not provide superior vaccine efficacy when children aged 5–17 months are vaccinated according to the standard M012 schedule. Nevertheless, during the 20-month period of this interim analysis, vaccine efficacy for each RTS,S group versus control was consistent with efficacy results from previous clinical trials.^2,3,19

Our findings differ from previous observations from a CHMI trial in malaria-naive adults, which suggested improved efficacy of the vaccine with fractional dosing.⁷ However, these CHMI studies evaluated vaccine efficacy against P falciparum infection following homologous challenge with high-density sporozoite infected laboratory-raised mosquitos.^6,7,20 This vaccine efficacy estimate cannot fully translate into efficacy against clinical disease in field settings where exposure might be heterologous and is the result of multiple bites from wild mosquitoes with potentially lower sporozoite density infections. In a trial in 6–10-week-old infants from Ghana, Tanzania, and Gabon, when delaying the administration of the third full dose from month 2 to month 7, no efficacy was observed before the third dose, and no improvement of vaccine efficacy against clinical malaria was noted over a period of 19 months of follow-up from first dose.²¹

Moreover, it was previously hypothesised, based on germinal-centre and B-cell biology,²² that in the context of challenge studies in malaria-naive adults, the use of a fractional instead of a full RTS,S dose might have enhanced competitive antigen binding in germinal centres. This would have led to preferential survival and expansion of circumsporozoite-specific B cells with the highest antigen affinity and to the higher antibody avidity observed in CHMI trials. However, such improved avidity was not observed in the current study, possibly because in children constantly exposed to malaria parasites through natural infection, B-cell affinity maturation might be altered,²³ resulting in the loss of the advantage of the fractional dose regimen seen in CHMI studies.

Overall, over the full 20-month follow-up period, substantial protection was observed in all RTS,S groups across the various endpoints assessed. The lowest vaccine efficacy was seen in the R012–20 group (in which the fourth dose given at month 20 did not contribute to the assessed efficacy as it was given at the end of the follow-up period). However, this was not the case over the initial 12 months of follow-up after dose three, when vaccine efficacy point estimates were very similar across groups. Interestingly, vaccine efficacy in the delayed fractional-dose group (Fx017–20) was similar to that in the other groups despite the 5-month delay in the administration of the third dose and the fact that the fourth dose (given at month 20) also did not contribute to the assessed efficacy. This might indicate either that the fractional RTS,S dose can be delayed by 5 months without losing protection against malaria, or that the initial lower protection in the Fx017–20 group after dose two is compensated by a higher efficacy of the delayed third fractional dose.

We observed a difference in the background incidence of clinical malaria between Ghana and Kenya, similar to previous reports for malaria incidence in areas close to the study sites.²⁴ A trend for lower vaccine efficacy of RTS,S with higher malaria incidence was previously described, but a statistical proof of an interaction between vaccine efficacy and transmission intensity could not be established.^3,19 In our study, we also observed higher point estimates of efficacy with lower background malaria incidence in Ghana (except in the Fx012–14 group) over the 20-month follow-up period. However, our study was not designed to assess differences between countries or country-specific data, thus observed differences could also be due to chance.

Over the initial 12 months of follow-up after dose three, there was a discrepancy between groups receiving the same full-dose regimen (R012–20 and R012–14) when assessing vaccine efficacy against all episodes of clinical malaria. This discrepancy is driven by results from Kenya. However, a post-hoc analysis showed that, during this period, the incidence of clinical malaria episodes did not differ significantly between the R012–20 and R012–14 groups, although point estimates were lower in the R012–14 group than the R012–20 group. Therefore, we concluded that this was most probably a chance finding, which could nevertheless have contributed to the overall efficacy over 20 months of follow-up. Therefore, any conclusion on the potential benefit of a fourth dose given 12 months after dose three compared with 18 months after dose three should be drawn with caution this early in the study.

Anti-circumsporozoite protein antibody responses to RTS,S were similar between groups. Consistent with previous observations, the fourth dose did not boost anti-circumsporozoite protein antibodies to higher concentrations than those achieved after dose three.^3,21 This is in contrast to the anti-HBs responses that show an incremental response after each dose, with booster responses largely exceeding the antibody concentration after dose three.

In this study, we did not observe a difference in antibody avidity between the fractional-dose and full-dose regimens. In a previous CHMI study, the delayed fractional dose was shown to increase circumsporozoite protein-specific antibody avidity, which was hypothesised to be a potential contributor to the improved vaccine efficacy of an Fx017 regimen over an R012 regimen.^7,25 However, in a more recent CHMI challenge trial, anti-circumsporozoite protein-specific antibody avidity was not different between protected and unprotected individuals.⁶ The avidity of anti-NANP IgG responses elicited by RTS,S administered according to full-dose R012 and R017 regimens was not shown to be associated with protection from clinical malaria in children enrolled in two phase 2 field trials.^26,27 By contrast, anti-circumsporozoite protein IgG concentrations and avidity, which varied with age, site, and prevaccination concentrations, were shown to contribute to protection against clinical malaria in the RTS,S phase 3 efficacy trial.²⁸

All regimens were well tolerated, supporting the acceptable safety profile of RTS,S. The incidence of solicited adverse events was similar to that observed in the phase 3 trial assessing the full-dose M012 regimen in African children aged 5–17 months.² Meningitis and cerebral malaria were previously highlighted as safety signals in the RTS,S phase 3 trial.^3,29 We did not observe any indication of increased risk for suspected or cause-confirmed meningitis in the RTS,S groups compared with the control group, and the only case of cerebral malaria was reported in the control group. This is in line with recent findings from the ongoing pilot implementation of RTS,S.³⁰

Our study has several limitations. Because the primary objective was not demonstrated, any further group comparisons should be interpreted with caution. Comparison between the delayed fractional regimen and groups receiving a M012 schedule might be hindered by seasonal variations in malaria incidence and transmission; in addition, the study did not include a R017–20 full-dose group to allow a non-biased comparison with the fractional Fx017–20 regimen. The study was only powered for overall results and not by site; this, together with the heterogeneity observed for vaccine efficacy estimates, known difference in malaria transmission even within the sites, and the multiplicity of objectives, limit the interpretation of results by country. Due to the open-label design, some bias in safety assessments is possible, although the study was designed to minimise it (ie, vaccinations and safety assessments were done by different individuals).

In conclusion, the Fx012–14 regimen did not show superior vaccine efficacy against clinical malaria after the three first doses when compared with a standard full-dose M012 schedule, and improved vaccine efficacy was also not observed for the delayed Fx017–20 regimen. However, all fractional-dose and full-dose RTS,S regimens provided substantial and similar protection against malaria.

Our results suggest that the use of a delayed fractional-dose regimen does not affect protective efficacy over 20 months of follow-up compared with a standard regimen, and that the timing of the third and fourth RTS,S vaccinations might be flexible. Although not confirmatory, if borne out over the 50-month follow-up, these findings might portend substantial public health benefit, as reduced vaccine volume requirements and flexibility in dose timing would allow increased access to the vaccine and decreased malaria-specific morbidity and mortality. Continued follow-up to month 50 will provide further insight into the efficacy of different RTS,S regimens, including the effect of multiple annual booster doses.