Abstract

METHODS

STUDY DESIGN AND SEQUENCE-DATA GENERATION

The RTS,S phase 3 trial design has been described previously (for details, see ClinicalTrials.gov number NCT00866619).^4-7 Samples were analyzed from vaccine recipients at all 11 trial sites, spanning seven countries in Africa (Fig. 1A). Samples representing two protocol-specified end points were sequenced in the per-protocol cohort (i.e., participants who received all three vaccinations at months 0, 1, and 2): primary clinical malaria (first or only episode of clinical malaria with a parasite density >5000 per cubic millimeter), occurring between 14 and 385 days after the third vaccination, and parasite positivity (parasite density >0 per cubic millimeter, regardless of whether symptoms were present), occurring at 18 months after vaccination.³

Open in a separate window

Figure 1

Study Sites and Genomic Units.

Samples from participants were received as dried blood spots on Whatman FTA sample cards. The methods we used for DNA extraction, PCR amplification, and sequencing are described in the Supplementary Appendix. The circumsporozoite protein C-terminal and SERA-2 amplicons were sequenced on an Illumina MiSeq platform. The NANP–NVDP amplicon was sequenced on a PacBio platform because of its greater length. Fig. 1A shows the respective locations of the NANP–NVDP repeat region and C-terminal amplicon within the circumsporozoite protein. The analyses included MiSeq data for 4421 circumsporozoite protein C-terminal samples and 4499 SERA-2 samples and PacBio data for 3137 circumsporozoite protein NANP–NVDP samples (Tables S1 through S4 and the statistical analysis plan in the Supplementary Appendix). All MiSeq and PacBio amplicon sequence data were submitted to the NCBI Sequence Read Archive³⁵ (BioProject PRJNA235895).

RESULTS

PARTICIPANTS AND SAMPLES

Fig. 2, and Fig. S2 in the Supplementary Appendix, summarize sample size and follow-up information for the per-protocol group of participants who were 5 to 17 months of age; in this cohort, P. falciparum genetic data were analyzed from 1181 RTS,S/AS01-vaccine recipients and 909 control-vaccine recipients in whom the primary clinical malaria end point was confirmed, as well as from 284 RTS,S/AS01-vaccine recipients and 208 control-vaccine recipients in whom the parasite-positivity end point was confirmed. Fig. S3 and S4 in the Supplementary Appendix show this information for the per-protocol group of infants who were 6 to 12 weeks of age, and Fig. S5 through S8 in the Supplementary Appendix provide information on the samples studied for both end points and age categories for the NANP–NVDP repeat amplicon.

Open in a separate window

Figure 2

Data Generation and Sample and Data Filtration for the End Point of Primary Clinical Malaria among Children 5 to 17 Months of Age

Vaccine recipients were considered to be out of interval for the dose regimen if they did not receive booster vaccinations according to the schedule specified by the trial protocol. The sample cards used were Whatman FTA cards. PCR denotes polymerase chain reaction.

COMPLEXITY OF INFECTION

The majority of samples from participants with primary clinical malaria in both age categories (68% of samples from infants 6 to 12 weeks of age and 65% of samples from children 5 to 17 months of age) had complex infections, defined as being founded by two or more distinct parasite lineages. In the older age category, the distribution of complexity of infection was shifted toward fewer parasite lineages in RTS,S/AS01-vaccine recipients than in control-vaccine recipients (complex infections in 61% vs. 71% of samples, P<0.001 by Wald test) (Fig. 3A), whereas in the younger age category there was no evidence of a different distribution between the RTS,S/AS01 and control groups (67% and 70%, respectively; P = 0.43) (Fig. S10 in the Supplementary Appendix). This observation for the older age category is concordant with findings in two phase 2 trials of the related RTS,S/AS02 vaccine,^31,32 and there were fewer 3D7-matching full-amplicon haplotypes in infections of high complexity in the RTS,S/AS01-vaccinated group than in the control group (Fig. S11 in the Supplementary Appendix).

Open in a separate window

Figure 3

Complexity of Infection and Frequencies of 3D7-Matched Malaria among Children 5 to 17 Months of Age with the Primary Clinical Malaria End Point

The denominator for the frequencies in Panels B through D is the number of samples representing the primary clinical malaria end point that had sequence data available. Ag denotes Agogo, Ba Bagamoyo, Kil Kilifi, Kin Kintampo, Kom Kombewa, Kor Korogwe, La Lambaréné, Li Lilongwe, Na Nanoro, and Si Siaya.

C-TERMINAL REGION SIEVE EFFECTS

Through 1 year after vaccination, we detected 139 clinical malaria cases with a perfect full-amplicon 3D7 match (Fig. 4A) and 1951 cases that were mismatched (Fig. 4B). During this period, the cumulative vaccine efficacy against clinical malaria with a perfect full-amplicon 3D7 match was 50.3% (95% confidence interval [CI], 34.6 to 62.3), and that against mismatched clinical malaria was 33.4% (95% CI, 29.3 to 37.2), with vaccine efficacy significantly higher against matched malaria (P = 0.04 by Wald test for the sieve effect) (Fig. 4D and Fig. 5A). The covariate-adjusted analysis gave almost identical results (Table S5 in the Supplementary Appendix). The cumulative vaccine efficacy was higher against matched malaria than against mismatched malaria throughout the follow-up period; for example, during the period through month 6, vaccine efficacy against matched malaria was 70.2% (95% CI, 56.1 to 79.8), and that against mismatched malaria was 56.3% (95% CI, 51.1 to 60.9) (P = 0.05 for the sieve effect) (Fig. 4C). Cumulative vaccine efficacy and sieve effects for the circumsporozoite protein C-terminal also varied in magnitude among study sites when the sites were analyzed individually in the older age category (Table S6 in the Supplementary Appendix).

Open in a separate window

Figure 4

Cumulative Incidences and Vaccine Efficacy against the Primary Clinical Malaria End Point among Children 5 to 17 Months of Age

Shown is the cumulative incidence of 3D7-matched malaria (Panel A) and 3D7-mismatched malaria (Panel B) among RTS,S/AS01-vaccine recipients and control-vaccine recipients during 12 months of post-vaccination follow-up, the cumulative vaccine efficacy (one minus the ratio [RTS,S/AS01 vs. control] of cumulative incidences of the first or only episode of clinical malaria with a specific haplotype) against 3D7-matched and 3D7-mismatched malaria over the entire post-vaccination follow-up period (Panel C), and the cumulative vaccine efficacy and hazard-ratio vaccine efficacy (one minus the ratio [RTS,S/AS01 vs. control] of instantaneous incidences of the end point under the assumption that incidences are proportional over time) against 3D7-matched and 3D7-mismatched malaria at 12 months after vaccination, with tests for differential haplotype-specific vaccine efficacy (Panel D). The I bars in Panel D indicate 95% confidence intervals.

Open in a separate window

Figure 5

Vaccine Efficacy against the Primary Clinical Malaria End Point among Children 5 to 17 Months of Age

Shown are the cumulative vaccine efficacy (Panel A) and hazard-ratio vaccine efficacy (Panel B) for the prevention of clinical malaria in which parasites were matches or mismatches with the RTS,S/AS01 vaccine strain (3D7) at each haplotype locus. Estimates were stratified according to study site. For each haplotype locus, the calculation of haplotype-matched vaccine efficacy included only clinical malaria end-point events with samples in which parasites matched 3D7 at the given locus; the calculation of haplotype-mismatched vaccine efficacy included only clinical malaria end-point events with samples in which parasites mismatched 3D7 at the given locus. Asterisks indicate that the difference in efficacy was significant (Q value ≤0.2 for all 28 multiply compared haplotype loci and unadjusted P≤0.05 for the full circumsporozoite protein C-terminal amplicon). FWER denotes family-wise error rate.

The hazard-ratio vaccine efficacy over the 12 months of post-vaccination follow-up was also higher against 3D7-matched malaria than against 3D7-mismatched malaria (62.7% [95% CI, 51.6 to 71.3] vs. 54.2% [95% CI, 49.9 to 58.1]; P = 0.06 for the sieve effect) (Fig. 5B, and Table S7 in the Supplementary Appendix). The overall vaccine efficacy was similar to the efficacy against mismatched malaria, because more than 90% of the infections were mismatched (Fig. 4D). Post hoc analysis defining the haplotypes of a malaria case as “any match” or “no match” to 3D7 also identified a sieve effect for the circumsporozoite protein C-terminal (P<0.001 for the cumulative vaccine efficacy sieve effect; P = 0.002 for the hazard-ratio vaccine efficacy sieve effect) (Tables S8 and S9 in the Supplementary Appendix). In contrast, the cumulative and hazard-ratio vaccine efficacies were similar against full-amplicon–matched malaria and full-amplicon–mismatched malaria in the category of infants 6 to 12 weeks of age (P = 0.58 for the sieve effect) (Tables S10 and S11 and Fig. S12 in the Supplementary Appendix), as well as against SERA-2– matched malaria and SERA-2–mismatched malaria in both age categories (P values for the sieve effect >0.30) (Tables S12 through S15 and Fig. S13 and S14 in the Supplementary Appendix).

Among the participants who were 5 to 17 months of age, there were also significant cumulative vaccine efficacy sieve effects (Q value ≤0.2) for the Th2R and Th3R epitopes, the Th2R–Th3R LD haplotype, and the DV10 region (Fig. 5A), as well as for the individual amino acid positions 299, 301, 317, 354, 356, 359, and 361 (Table 1). The cumulative vaccine efficacy tended to decrease with the number of mismatches with 3D7 at these seven amino acid positions (Fig. S15 in the Supplementary Appendix). Hazard-ratio analyses of epitopes and regions (Fig. 5B) and individual amino acid positions (Table S7 in the Supplementary Appendix) yielded differential vaccine efficacy results that were consistent with those from the cumulative vaccine efficacy analysis, at reduced levels of significance. In the younger age category, vaccine efficacy against malaria that matched 3D7 at individual circumsporozoite protein C-terminal amino acid positions was similar to that against malaria with mismatches (all Q values >0.20) (Tables S10 and S11 in the Supplementary Appendix). No evidence of sieve effects was found for individual positions in either age category for the SERA-2 locus (Table S12 through S15 in the Supplementary Appendix).

Table 1

Cumulative Vaccine Efficacy against Primary Clinical Malaria.^*

Locus and Amino Acid Position†	3D7-Matched Malaria‡		3D7-Mismatched Malaria‡		Differential Efficacy
	Vaccine Efficacy (95% CI)	P Value	Vaccine Efficacy (95% CI)	P Value	P Value	FWER P Value§	Q Value
	percent		percent
DV10
294	34.8 (30.8 to 38.6)	<0.001	31.3 (−6.4 to 55.6)	0.09	0.83	1.00	0.89
295	34.9 (31.0 to 38.6)	<0.001	12.6 (−78.8 to 57.2)	0.71	0.44	1.00	0.71
296	34.8 (30.9 to 38.5)	<0.001	7.1 (−310.9 to 79.0)	0.92	0.67	1.00	0.89
298	33.5 (29.1 to 37.7)	<0.001	41.7 (30.1 to 51.4)	<0.001	0.20	1.00	0.41
299¶	35.4 (31.5 to 39.1)	<0.001	−62.8 (−186.0 to 7.3)	0.09	0.003	0.08	0.08
301¶	49.2 (35.3 to 60.1)	<0.001	33.1 (28.9 to 37.0)	<0.001	0.03	0.81	0.15
302	34.8 (30.9 to 38.5)	<0.001	20.0 (−108.2 to 69.3)	0.64	0.69	1.00	0.89
303	34.6 (30.7 to 38.3)	<0.001	68.4 (−100.9 to 95.0)	0.22	0.45	1.00	0.71
Th2R
314	32.7 (26.2 to 38.2)	<0.001	37.3 (30.9 to 43.0)	<0.001	0.33	1.00	0.61
317¶	45.9 (33.3 to 56.2)	<0.001	33.1 (28.9 to 37.1)	<0.001	0.06	1.00	0.19
318	36.9 (25.8 to 46.4)	<0.001	34.2 (29.8 to 38.4)	<0.001	0.65	1.00	0.89
320	34.4 (30.4 to 38.1)	<0.001	64.0 (21.0 to 83.6)	0.01	0.14	1.00	0.32
321	35.7 (24.3 to 45.3)	<0.001	34.5 (30.1 to 38.7)	<0.001	0.85	1.00	0.89
322	34.8 (27.2 to 41.5)	<0.001	34.7 (29.4 to 39.6)	<0.001	0.99	1.00	0.99
324	37.3 (32.0 to 42.2)	<0.001	30.7 (23.3 to 37.4)	<0.001	0.16	1.00	0.35
327	35.5 (31.2 to 39.5)	<0.001	30.1 (16.5 to 41.4)	<0.001	0.42	1.00	0.71
Th3R
349	34.8 (30.9 to 38.5)	<0.001	25.4 (−84.1 to 69.8)	0.53	0.81	1.00	0.89
352	35.1 (30.7 to 39.2)	<0.001	32.8 (20.0 to 43.5)	<0.001	0.73	1.00	0.89
354¶	36.0 (32.0 to 39.8)	<0.001	10.8 (−22.6 to 35.1)	0.48	0.05	1.00	0.17
355	34.7 (30.8 to 38.4)	<0.001	54.1 (−286.1 to 94.5)	0.47	0.71	1.00	0.89
356¶	36.2 (32.2 to 40.1)	<0.001	15.8 (−7.8 to 34.3)	0.17	0.04	0.85	0.15
357	35.3 (27.3 to 42.5)	<0.001	34.4 (29.3 to 39.2)	<0.001	0.86	1.00	0.89
359¶	36.1 (32.0 to 40.1)	<0.001	22.2 (5.2 to 36.2)	0.01	0.07	1.00	0.20
361¶	39.3 (33.6 to 44.4)	<0.001	29.3 (22.4 to 35.5)	<0.001	0.03	0.81	0.15

Open in a separate window

^*Data are for efficacy through 12 months after vaccination among children 5 to 17 months of age. Estimates were stratified by study site. CI denotes confidence interval.

^†Only circumsporozoite protein C-terminal amino acid positions with a sufficiently high minor-allele frequency were included in the analysis.

^‡For each amino acid position, the calculation of haplotype-matched vaccine efficacy included only clinical malaria end-point events with samples in which parasites matched 3D7 at the given position; the calculation of haplotype-mismatched vaccine efficacy included only clinical malaria end-point events with samples in which parasites mismatched 3D7 at the given position.

^§The family-wise error rate (FWER) P value is an adjusted³⁹ P value controlling for the FWER.

^¶The differential efficacy was significant; significance was defined as Q value of up to 0.2 for all 28 multiply compared haplotype loci (all epitope regions and amino acid positions with a sufficiently high minor-allele frequency).

The linkage disequilibrium (LD) haplotype includes this amino acid position.

For the parasite-positivity end point at 18 months after dose 3, in the older age category, the estimates of vaccine efficacy tended to be higher for circumsporozoite protein C-terminal 3D7-matched malaria than for 3D7-mismatched malaria (e.g., vaccine efficacy, 53% vs. 30%; P = 0.19 for the full amplicon), although none of the differences were significant (Table S16 in the Supplementary Appendix). In contrast, there was no evidence at all of a sieve effect for the circumsporozoite protein C-terminal in the younger age category with regard to this end point (Table S17 in the Supplementary Appendix), and there was no evidence of a sieve effect for SERA-2 in either age category (Tables S18 and S19 in the Supplementary Appendix).

DISCUSSION

The discovery that RTS,S/AS01 vaccine efficacy is higher against clinical malaria with infections matching the 3D7 vaccine construct at epitope haplotypes and amino acid positions than it is against infections not matching 3D7 is not entirely unexpected, given the polymorphism at the C-terminal of the circumsporozoite protein antigenic locus and previous observations of allele-specific immune responses to other parasite proteins.^27,29 This differential cumulative vaccine efficacy result could be a false positive result, given that the result for the full circumsporozoite protein C-terminal was of borderline significance (P = 0.04) and that the analysis of four epitopes and 24 amino acid sites gave no Holm–Bonferroni-adjusted P values below 0.05. However, 11 of the 28 tests yielded Q values less than or equal to 0.20; consequently, we expect at least 80% of these 11 results to be true positives.

The current study had greater power to detect allele-specific protection than did previous evaluations of RTS,S in phase 2 trials, because of three factors: a larger sample, the inclusion of study sites harboring a higher frequency of 3D7-matching haplotypes, and the use of PCR-based next-generation sequencing to resolve the haplotypes that make up mixed infections. Our main result of significant sieve effects for the primary clinical malaria end point in the age category of 5 to 17 months was based on a large sample, with measured genetic data from 2145 total clinical malaria cases. In contrast, the sieve analysis of the parasite-positivity end point in participants 5 to 17 months of age, which yielded nonsignificant results, had many fewer cases (genetic data from 507 total cases). However, in terms of estimates of vaccine efficacy, the sieve effects were slightly stronger for the parasite-positivity end point, which suggests that the lack of significance could be due to lower statistical power rather than to a true lack of differential protection. Post hoc power calculations showed only 30% power to detect the 23-percentage-point difference between vaccine efficacies (53% against malaria matching the 3D7 circumsporozoite protein C-terminal vs. 30% against mismatched malaria) with regard to the parasite-positivity end point, as compared with 51% power to detect such a difference with regard to the clinical malaria end point. Therefore, the selective vaccine protection may have operated for both clinical malaria and parasite positivity. In contrast, there was no evidence for allele-specific vaccine efficacy in the age category of 6 to 12 weeks, despite the significant over all protection in this group. This result implies a qualitative difference in the vaccine response, in addition to the previously reported quantitative difference in anti-circumsporozoite antibody titers in the younger age category.⁴ The biologic mechanisms underlying the differences between the age categories remain to be elucidated, but they could include a role of maternal antibody, interactions with other vaccine responses, or differences in immune-response capacity between infants and children. Further immunologic analyses could clarify the mechanisms of selective versus nonselective vaccine-induced immunity.

Genetic surveillance of circumsporozoite protein sequences in parasite populations could inform the development of future vaccine candidates targeting polymorphic malaria parasite proteins. The genotype-specific vaccine efficacy results we report here complement previous estimates of RTS,S/AS01 efficacy in this phase 3 trial; the previously reported 12-month hazard-ratio vaccine efficacy of 55.8% against clinical malaria among children 5 to 17 months of age⁴ can now be interpreted as a multiplicative weighted mean of hazard-ratio vaccine efficacy values of 62.7% against matched parasites (139 infections) and 54.2% against mismatched parasites (1951 infections) (Fig. 5B). The observed variation among study sites in infections with a perfect vaccine match in the circumsporozoite protein C-terminal (Fig. 3B, and Fig. S18 and S19 in the Supplementary Appendix) may help to explain previously reported variation in overall vaccine efficacy among study sites, although the magnitude of this contribution is expected to be low because of the overall rarity of the 3D7 haplotype.⁶ Broader deployment of the vaccine could result in increased selection on the 3D7 haplotype or its component epitopes and amino acid alleles. Sieve analysis of next-generation sequencing data constitutes an approach for understanding partial vaccine efficacy.

Supplementary Material

Acknowledgments

APPENDIX

The authors’ full names and academic degrees are as follows: Daniel E. Neafsey, Ph.D., Michal Juraska, Ph.D., Trevor Bedford, Ph.D., David Benkeser, M.P.H., Clarissa Valim, M.D., Sc.D., Allison Griggs, M.Sc., Marc Lievens, M.Sc., Salim Abdulla, M.D., Ph.D., Samuel Adjei, M.B., Ch.B., D.T.M., Tsiri Agbenyega, M.B., Ch.B., Ph.D., Selidji T. Agnandji, M.D., Pedro Aide, M.D., Ph.D., Scott Anderson, B.S., Daniel Ansong, M.B., Ch.B., John J. Aponte, M.D., Ph.D., Kwaku Poku Asante, M.D., Ph.D., Philip Bejon, Ph.D., Ashley J. Birkett, Ph.D., Myriam Bruls, M.Sc., Kristen M. Connolly, B.S., Umberto D’Alessandro, M.D., Ph.D., Carlota Dobaño, Ph.D., Samwel Gesase, M.D., Brian Greenwood, M.D., Jonna Grimsby, Ph.D., Halidou Tinto, Ph.D., Mary J. Hamel, M.D., D.T.M.&H., Irving Hoffman, P.A., M.P.H., Portia Kamthunzi, M.D., Simon Kariuki, Ph.D., Peter G. Kremsner, M.D., Amanda Leach, M.R.C.P.C.H., Bertrand Lell, M.D., Niall J. Lennon, Ph.D., John Lusingu, M.D., Ph.D., Kevin Marsh, M.D., Francis Martinson, M.D., Ph.D., Jackson T. Molel, M.Sc., Eli L. Moss, B.S., Patricia Njuguna, M.B., Ch.B., M.Med., Christian F. Ockenhouse, M.D., Ph.D., Bernhards Ragama Ogutu, Ph.D, Walter Otieno, M.B., Ch.B., M.Med., Lucas Otieno, M.B., Ch.B., M.P.H., Kephas Otieno, M.P.H., Seth Owusu-Agyei, Ph.D., Daniel J. Park, Ph.D., Karell Pellé, Ph.D., Dana Robbins, B.S., Carsten Russ, Ph.D., Elizabeth M. Ryan, B.S., Jahit Sacarlal, M.D., Ph.D., Brian Sogoloff, B.S., Hermann Sorgho, Ph.D., Marcel Tanner, Ph.D., Thor Theander, M.D., D.Sc., Innocent Valea, Pharm.D., Ph.D., Sarah K. Volkman, Sc.D., Qing Yu, M.Sc., Didier Lapierre, M.D., Bruce W. Birren, Ph.D., Peter B. Gilbert, Ph.D., and Dyann F. Wirth, Ph.D.

The authors’ affiliations are as follows: the Broad Institute of Massachusetts Institute of Technology and Harvard (D.E.N., C.V., A.G., S. Anderson, K.M.C., J.G., N.J.L., E.L.M., D.J.P., K.P., D.R., C.R., E.M.R., B.S., S.K.V., Q.Y., B.W.B.), Cambridge, and Harvard T.H. Chan School of Public Health (C.V., K.P., S.K.V., D.F.W.) and Simmons College School of Nursing and Health Sciences (S.K.V.), Boston — all in Massachusetts; Fred Hutchinson Cancer Research Center (M.J., T.B., D.B., C.V., P.B.G.) and University of Washington (D.B., P.B.G.), Seattle; GlaxoSmithKline Vaccines, Rixensart, Belgium (M.L., M.B., A.L., D.L.); Ifakara Health Institute, Bagamoyo (S. Abdulla, J.T.M.), and National Institute for Medical Research, Korogwe (S.G., J.L.) — both in Tanzania; School of Medical Sciences, Kwame Nkrumah University of Science and Technology, Kumasi (S. Adjei, T.A.), and Kintampo Health Research Centre, Kintampo (K.P.A., B.G., S.O.-A.) — both in Ghana; Albert Schweitzer Hospital, Lambaréné, Gabon (S.T.A., D.A., P.G.K., B.L.); Institute of Tropical Medicine, University of Tübingen, Tübingen, Germany (S.T.A., P.G.K., B.L.); Centro de Investigaçåo em Saùde de Manhiça, Manhiça, Mozambique (P.A., J.J.A., C.D., J.S.); ISGlobal, Barcelona Center for International Health Research, Hospital Clinic–Universitat de Barcelona, Barcelona (P.A., J.J.A., C.D.); PATH Malaria Vaccine Initiative, Washington, DC (A.J.B., C.F.O.); Medical Research Council Unit, Banjul, Gambia (U.D.); London School of Hygiene and Tropical Medicine, London (B.G.); Institut de Recherche en Sciences de la Santé, Nanoro, Burkina Faso (H.T., H.S., I.V.); Centers for Disease Control and Prevention (CDC), Atlanta (M.J.H.); University of North Carolina (UNC), Chapel Hill (I.H.); UNC Project, Lilongwe, Malawi (P.K., F.M.); Kenya Medical Research Institute (KEMRI)–Wellcome Trust Research Program, Kilifi (P.B., K.M., P.N.), KEMRI–CDC Research and Public Health Collaboration, Kisumu (S.K., K.O.), and KEMRI–Walter Reed Project, Kombewa (B.R.O., W.O., L.O., K.P.) — all in Kenya; Swiss Tropical and Public Health Institute, Basel, Switzerland (M.T.); and the University of Copenhagen, Copenhagen (T.T.).

Footnotes

REFERENCES