Skip to main content
European Commission logo print header

Identifying correlates of protection to accelerate vaccine trials: systems evaluation of two models of experimentally-induced immunity to malaria

Final Report Summary - SYSMALVAC (Identifying correlates of protection to accelerate vaccine trials: systems evaluation of two models of experimentally-induced immunity to malaria)

Executive Summary:
Executive summary
Clinical development of vaccines against infectious diseases is a lengthy and costly process, and particularly challenging for those diseases for which immune responses conferring protective immunity are unknown. Malaria, an acute febrile illness often causing fatality in children and infants, is one clear example of a complex disease for which despite several years of research there is uncertainty about what immune responses need to be induced in order to prevent it. For this reason, current experimental malaria vaccines have largely been developed applying empirical (trial & error) strategies, and thus far no clear correlation has been established between immunogenicity and protection. Where classical reductionist approaches have failed to identify immune correlates of protection, the integration of omics technologies and modern systems biology approaches have the greatest potential to significantly advance in the field of vaccinology. The identification of in vitro surrogate markers of protection and the understanding of the mechanisms conferring protection against malaria is key to the rational design of future more effective vaccines, which will significantly accelerate the development of new or improved candidates by reducing the size, length and expense of clinical trials.
The SysMalVac project, entitled “Identifying correlates of protection to accelerate vaccine trials: systems evaluation of two models of experimentally-induced immunity to malaria” started in 2013 in the context of the vaccine immunology studies conducted within the multicentre Phase III clinical trial of RTS,S/AS01E in African children. RTS,S, a subunit malaria vaccine candidate, is the most advanced clinically and one of the very few that have shown consistent protection in several field trials. However, its mode of action remains only partially known. In addition, other immunisation strategies developed in Europe based on whole live parasites, like the chloroquine chemoprophylaxis with Plasmodium falciparum sporozoites (CPS) approach, have been able to induce sterile immunity experimentally. The mechanism of protection remains also fragmentary. This project brought together teams of researchers working on these two immunisation approaches that have been able to induce protection against malaria both in the field and under experimental challenge conditions. The aim is to identify common and unique immune correlates underlying protection by applying novel systems biology approaches in partnership with European industrial partners.
The goal of SysMalVac is to use a systems biology analytical approach to interrogate data from two efficacious malaria vaccination models towards identifying combinatorial biomarkers of protection. Whole transcriptomic profiles and immunological read-outs were compared between human volunteers who were protected or not from malaria following immunisation. Data was integrated into an artificial intelligence-based analytical tool developed by Anaxomics that discriminated protein signatures that could predict protection or susceptibility with a high accuracy and generalisation capacity.
In addition, a non-human primate (NHP) model reproducing the CPS clinical trial was set up for the refinement and validation of the biomarkers identified from the human models. The NHP model allows for an in depth study of the immune system enabling the validation of the identified biomarkers by direct test of tissue resident effector cells from spleen and lymph nodes which are not accessible in human trials.
Finally, separate sets of samples from different individuals vaccinated with RTS,S or whole parasite immunisation strategies are used to validate the identified candidate biomarkers of protection.
In conclusion, the SysMalVac project demonstrates that the systems biology and modelling strategy applied to immunological and transcriptional data generated in the context of RTS,S and CPS clinical trials was successful in identifying and validating sets of combinatorial biomarkers predictive of whether an individual will be protected or not upon vaccination, with high accuracy and generalizability. This approach can be very valuable for optimising and accelerating vaccine trials and opens new avenues for identifying mechanisms of action of promising candidates.

Project Context and Objectives:
Summary description of project context and objectives
For the SysMalVac project, the following two efficacious malaria vaccination models were studied:
1. The RTS,S vaccine from GlaxoSmithKline Biologicals (GSK) which is currently under assessment for implementation in malaria-endemic countries after completion of the pivotal Phase III clinical trial. The RTS,S vaccine showed moderate but significant efficacy, however there is evidence of waning of the protective effect with time.
2. The chloroquine chemoprophylaxis with Plasmodium falciparum sporozoites (CPS) immunisation strategy based on controlled human malaria infections (CHMI) carried out in small groups of naïve volunteers exposed to bites of laboratory-reared mosquitoes infected with P. falciparum by artificial membrane feeding on gametocyte cultures. Safety and predictability of CHMI has been confirmed in numerous studies conducted in the USA and Europe. Recently CHMI studies have also been conducted in semi-immune volunteers in malaria endemic countries. RUMC demonstrated in a proof-of-concept study that sterile protection can efficiently be induced by repeated administration of sporozoites under chloroquine veil. This study potentially provides a rapid and effective strategy as a short cut to complete protection, serving as a blue print for acceleration of clinical malaria vaccine development.

High quality, well-characterised peripheral blood mononuclear cell (PBMC) samples were collected to investigate immunisation-induced immune responses from the RTS,S Phase III and CPS trials. RNA expression analysis was applied to perform a systems-wide transcriptional analysis of cellular responses upon parasite-specific antigenic re-stimulation.

Overall objective
To apply an analytical method of mapping the human immune response to RTS,S and CPS malaria immunisation strategies and allow a predictive artificial intelligence model to identify a biomarker signature correlating to protection. Common or generalizable in vitro immune correlates of protection will be refined and validated in an experimental CPS animal model as well as additional samples from human immunisation trials. The final validated set of transcriptional biomarkers will be bundled as a proposed surrogate of protection, which will be an actionable item for partners to develop a product work package.
Specific objectives
Aim 1: To use systems biology algorithms to identify a transcriptional signature of protection in two strategies of experimentally induced immunity against Plasmodium falciparum

Hypothesis: The mapping and artificial intelligence learning algorithm employed by AX will identify partially overlapping transcriptional profiles of in vitro immune responses in the protected groups in the RTS,S and CPS immunisation systems that are distinct from profiles in these individuals pre-immunisation and from responses in non-protected individuals.
Approach: For the RTS,S vaccine trial, pre-vaccination and post-vaccination cells from protected and non-protected African children were analysed for gene transcription after in vitro vaccine and parasite antigen re-stimulation. For the CPS immunisation trials, data was obtained from transcriptomics analyses of re-stimulated blood cell samples from adult volunteers who, under chloroquine prophylaxis, received P. falciparum infections (CPS, protected). Data was integrated and analysed for each immunisation strategy separately using the proposed immuno-network mapping and artificial intelligence model.
Outcome: To identify the overlapping transcriptional signature that predicts protection by RTS,S and CPS immunisation.

Aim 2: To refine and validate the algorithm and resulting transcriptional signature of protection using a non-human primate malaria model of experimentally induced immunity

Hypothesis: The protection achieved through CPS immunisation can be replicated in other immunisation models, as protective immunity is the result of similar activation and effector functions upon antigen recall responses in protected individuals.
Approach: Additional data obtained in a CPS experiment in non-human primates (NHP) was generated for parameterisation, refinement and testing of the hypothesis generated in the initial mathematical modelling (Aim 1). Using this animal model, biomarkers from the peripheral blood of both humans and NHP will be contrasted with those of lymphoid tissues and other organs, which cannot be measured in the human immunisation model but where the bulk of the immune response is expected to occur.
Outcome: To demonstrate the robustness of the analytical model and verify that the combinatorial biomarkers of protection found in the peripheral blood reflect the protective immune response expected to occur in the lymphoid and malaria-affected tissues (e.g. liver). This will provide additional support for the common biomarkers that predict protective efficacy in the two immunisation strategies.

Aim 3: To validate a biomarker signature that will potentiate development of an optimal tool for vaccine candidate evaluation

Hypothesis: The identified transcriptional biomarker signature (Aim 2) can be validated to reproducibly predict protection in CPS- or RTS,S-immunised individuals.
Approach: Combinatorial biomarkers of protection were validated using existing biological samples from human trials independent from the samples used to generate the model.
Outcome: Validated transcriptional biomarker as a potential surrogate marker of protection. An actionable recommendation will be made for the development of a tool or product work package for vaccine selection and evaluation without the need for large efficacy trials.

Project Results:
Description of the main S&T results/foregrounds
The project was divided into Work Packages (WP). Results are detailed below as per the tasks carried out in each WP.
WP1 - Data generation and management with existing samples from human trials
The main objective of this WP is to generate transcriptomic data from PBMC samples obtained from the CPS and RTS,S immunisation models, previously stimulated with parasite antigen. RNA purified from the stimulated PBMCs were subjected to expression analysis by microarray and the identified transcripts involved in the immunisation responses induced by RTS,S and CPS immunization strategies were integrated into the mathematical model in WP2.

Gathering and management of existing data from human trials
Clinical, parasitological and protection data were compiled for both CPS and RTS,S studies.

Antigen stimulation of PBMC samples from human trials for RNA isolation and supernatant Luminex analysis
Standard operating procedures (SOP) for both in vitro re-stimulation experiments and flow cytometry analysis were developed. Preliminary experiments to optimize and harmonize stimulation conditions were carried out with samples from a separate CPS trial. Optimal stimulation conditions were employed in SysMalVac based on RNA profiling data and confirmed by Luminex analysis of supernatants obtained in the pilot CPS study. In addition, a flow cytometry phenotyping panel was optimized to quantify proportions of different cell types before antigen stimulation to complement RNA expression profiling. In vitro re-stimulation experiments of pre- and post-immunization PBMC samples from CPS and RTS,S trials were performed to obtain PBMC pellets for RNA expression profiling and cell culture supernatants for analysis of secreted biomarkers. Whole transcriptomic profiles of RNA purified from PBMC pellets harvested after different antigen stimulations were analysed by Affimetrix microarrays. Cytokines, chemokines and growth factors were quantified in cell culture supernatants by Luminex. The proportions of the main blood cell leukocyte populations were phenotyped by flow cytometry.

Transcriptional data generation of stimulated PBMC
Optimisation studies, RNA isolation and QC checking were performed prior to processing samples for gene expression analysis by Affymetrix technology on GeneTitan. The obtained data was analysed by classical bioinformatics analyses to detect differentially expressed genes among protected and non protected individuals.

Results achieved
• Gathering of clinical and parasitological data from the CPS and RTS,S trials
• Successful preparation of stimuli (iRBC and uRBC), successfully optimising stimulation conditions, and harmonization of SOPs for cellular immunology studies.
• All cellular stimulation experiments on pre- and post-immunisation PBMCs of the CPS and RTS,S trials completed.
• Supernatants analysed for cytokines, chemokines and growth factors by Luminex analysis.
• RNA from samples isolated and quantified. Microarray platform run successfully and raw data results obtained for both classical bioinformatics analysis and integration in the mathematical model (WP2).

WP2 - Hypothesis generation and refinement: Computational modelling for biomarker signatures
Integration of all the data from the RTS,S and CPS trials together with data related to human immune responses available in the literature were gathered in order to generate mathematical models that provide a global view of the physiological processes related to protection in each immunization strategy.
The specific objectives are:
(1) Molecular characterization and creation of the protein network involved in human immune responses,
(2) Creation of mathematical models to identify main physiological processes leading to protection in the RTS,S vaccine and CPS immunization and
(3) to distil the models to identify combinatorial biomarkers of protection.

Molecular characterisation of human immune responses
The human immune response triggered by malaria was characterized at a molecular level through an exhaustive bibliographic research. The characterisation of malaria immune response resulted in the identification of 6 distinct motives: 1) Pre-erythrocytic stage immunity - Innate, 2) Pre-erythrocytic stage immunity - Adaptive, 3) Blood-stage immunity - Innate, 4) Blood-stage immunity - Adaptive, 5) Cellular receptors recognized by P. falciparum and 6) Immune response to skin-stage Plasmodium.
All malaria immune response motives were characterized at the protein level and 86 different proteins were identified and used for the network construction.

Creation of the protein network involved in human immune responses
The identified proteins were used as seeds to generate a map with a total of 1,715 proteins. The construction and analysis at topological level of the protein interaction network of human immune response triggered by malaria showed that the immune response was correctly characterized. Static, topological information is only the first step towards a complete understanding of any biological process; further construction of mathematical models helps to identify main physiological processes leading to protection in the RTS,S vaccine and CPS immunization.

Creation of mathematical models to identify the main physiological processes leading to protection in the RTS,S vaccine and CPS immunisation
The mathematical models have been created integrating gene expression data of PBMC. Prior to creating the mathematical models, the cell types that form the PBMCs were characterised at molecular level through an exhaustive literature review. Besides, other cell types that play a key role in malaria were characterised such as macrophages and the liver Kupffer cells. Then, the protein interaction network of human immune response triggered by malaria was modelled, obtaining mathematical models capable of both reproducing existing knowledge and predicting future data by using the TPMS technology. In addition, the TPMS technology was improved to integrate the specific information from expression microarrays in the mathematical models, allowing the construction of specific models that can be updated any time integrating the most up-to-date information. Thus, mathematical models capable of both reproducing existing knowledge and predicting future data were obtained.

RNA transcription data from the CPS model were integrated in the system obtaining mathematical models able to simulate the CPS immunization response. Specifically, the models have been created using all the available knowledge from databases and the literature included in the Anaxomics’ truth table, the hand-curated data that describes cellular functions, processes, etc, of PBMCs and the information extracted from gene expression arrays. This resulted in the construction of mathematical models with high accuracy.

The accuracy and uncertainty are statistical parameters that allow the evaluation of the constructed mathematical models. Accuracy refers to the percentage of fulfilment of the restrictions, the known information used to construct the mathematical model. In other words, how much the models comply with the scientific data used to train it.

The RTS,S RNA transcription data were firstly introduced into the mathematical model in a blinded fashion. This was done because the Phase III RTS,S study is a randomised double blinded trial and until the code was broken the vaccination assignment of the corresponding data was unknown. Thus, preliminary mathematical models for the RTS,S vaccine were generated by including the gene expression without knowing which individuals were vaccinated with RTS,S or with comparator vaccine. A preliminary list of RTS,S biomarker candidates was obtained for validation. Upon unblinding of the RTS,S trial by GSK, the final models were generated. These models also include the gene expression data provided by RNA microarrays experiments together with all the cellular functions and processes of PBMCs. Again, for this model, the constructed mathematical models were of high accuracy and thus able to reproduce a large part of the imposed restrictions.

The final conclusions of the mathematical models show high accuracy values. This fact indicates that the constructed mathematical models have high capability to reproduce the existing and the generated knowledge within the consortium.

Identification of biomarkers of protection for the RTS,S vaccine and CPS immunisation
All the cohorts generated for the CPS immunisation strategy and for the RTS,S vaccine were compared in pairs to identify the set of proteins that can take part in the best groups of proteins able to distinguish which individuals will be protected against malaria and which will not after the corresponding immunisation. These groups of proteins are known as classifiers.

A set of proteins that take part in the best classifiers (biomarkers) have been identified by the individual study of all CPS and RTS,S mathematical models. The best classifiers identified have shown accuracies between 70% and 100% and generalization capabilities measured by a leave-one-out cross-validation between 70% and 90%. Therefore, according to the results obtained, we have identified promising classifiers able to correctly predict if an individual will be protected or not against malaria by the CPS immunisation strategy or the RTS,S vaccine with high level of accuracy. Different sets of classifiers were obtained for each immunisation strategy, but there were some biomarkers that overlapped between the two.

Results achieved
• List of the “seed proteins” identified during the molecular characterization after RTS,S and CPS immunisations.
• Mathematical models able to reproduce the behaviour of the RTS,S and CPS immunisations.
• List of potential biomarkers to predict protection against malaria.

WP3 – Data generation with a non-human primate model
The main objective of this WP was to provide transcriptomics analysis of samples obtained from the Plasmodium knowlesi - Macaca mulatta malaria infection model. The identification of genes involved in this model will consolidate the transcriptomic results obtained by RTS,S and CPS trials.

The non-human primate (NHP) CPS study
The NHP CPS study uses the P. Knowlesi - M. mulatta malaria infection model to perform a sporozoite vaccination and challenge study following the protocols used for the CPS trials in humans. The transcriptomic analysis of samples obtained will be used to identify genes involved in protection thus allowing for the confirmation of the set of genes identified through transcriptomics in the human CPS and the RTS,S trials in an independent species.

To obtain the authorization to perform these animal studies, an application to the Animal Experimentation Committee was prepared and approval was granted. In order to obtain P. knowlesi sporozoites for vaccination and challenge it was necessary to gain access to a specific mosquito species able to transmit P. knowlesi sporozoites. This specific species is known as Anopheles dirus. For the whole duration of the study procedures were put in place for regular shipment of An. dirus eggs, which were successfully reared in the laboratory. Blood feeding of adult mosquitoes was also successful. Furthermore, protocols were harmonized to simulate the human CPS trial as closely as possible.

When the NHP study reached its end, contrary to the expected outcome, the immunised animals did not appear protected and both control and experimental animals became parasite positive after challenge. The Consortium analysed this result in depth and considered it possible that the immunisation sporozoite doses were too low. Therefore it was decided to cure the NHPs with antimalarials and proceed with a booster immunisation with higher sporozoite numbers under chloroquine veil and to re-challenge after a resting period allowing complete chloroquine washout.

The NHP study was continued with booster immunization and a new challenge. Sterile protection could at this point not be achieved. This may be due to several factors such as for example the ratio between immunisation and challenge dose, the route of challenge (intra venous versus mosquito bites in humans). Clearly, more work needs to be done to achieve sterile protection in this model. However, human CPS studies recently conducted seem to suggest that even when vaccinated individuals and controls do become patent in a similar timeframe, significant differences in the immune response between vaccinated and control individuals can be detected in the blood. Translated to our setting, this implies that even though the immune response may not suffice to bring about protection in the vaccinated animals, the immune response between the two groups may still be significantly different. In such a scenario, the objective of the SysMalVac study to evaluate the immune response in the periphery (blood) and correlate it to the tissue immune response (where the immune response is expected to take place) is therefore still scientifically meaningful. Furthermore, the information gained from the immune response in this setting, may help us better characterize the model and inform future studies in how to proceed to obtain protection.
Thus, ethical approval was obtained to follow the original protocol, and blood and tissue samples were obtained from all NHPs and samples were stimulated as per protocol.
The PBMC samples collected from the NHPs prior to vaccination and prior to challenge were in vitro re-stimulated with antigens. All procedures were carried out after standardisation with the human CPS protocols to keep differences with the human CPS study to a minimum.

All samples from re-stimulated PBMCs collected across the duration of the study were analysed by microarray analysis. The microarray data was analysed by classical bioinformatics analysis. The haematology and clinical chemistry parameters have also been integrated into the projects’ database.

Transcriptomics analysis in non-human primate blood samples
RNA was extracted from all samples, quantified and QC-assessed prior to performing the microarray analysis. The raw data has been integrated in the mathematical model.

Results achieved
• Selection of appropriate animals
• First time establishment of P. knowlesi sprozoite-based immunisation and challenge in Europe
• Performance of NHP study despite encountering a risk and contingency action
• PBMC samples re-stimulated in vitro with antigens and analysed by microarray analysis

The systems analysis of samples derived from the NHP study is currently being analysed and results will be published.
WP4 – Biomarker validation
The objective of this WP was to validate biomarker sets obtained in the hypothesis generation and biomarkers refinement phases (WP2) at the tissue level (NHP model) and in a new set of PBMC human samples by predicting the protection status of the individuals examined.

Validation of biomarkers with samples from non-human primate CPS trial
Pilot studies for the isolation of different cells from the liver of NHP were performed. Procedures to obtain cells from spleen and lymph nodes have been successfully tested. The characterisation of the isolated different liver cell populations based on surface markers has been performed.

At autopsy a selection of tissues (spleen, lymph nodes, liver) were obtained for immune cell isolation and subsequent in depth microarray analyses. Using these microarray data, immune responses from the peripheral blood of both humans and NHP will be compared with those of NHP lymphoid tissues and other organs, which cannot be measured in the human immunization models but where the bulk of the immune response is expected to occur. Due to the required extension of the NHP study, analysis is still on going.

Transcriptomic analysis in NHP organ samples
RNA isolation, quantification and QC testing was performed. Samples from all tissue types (spleen, liver, bone marrow and lymph node) was obtained and analysed. The samples were successfully hybridized on a monkey full genome expression microarray from Affymetrix and data has been integrated into the mathematical model.

Validation of biomarkers with separate sets of samples from human trials
An experimental validation of transcriptomic results by real time quantitative PCR analysis was performed.
The results of the experimental validation have shown that some of the proposed biomarkers could be good predictors of malaria protection forming classical linear classifiers. Furthermore, some of these linear classifiers only formed by two proteins that have showed 100% of accuracy.

The average of the separation capability of all the identified linear classifiers is 87%.

Non-linear classifiers able to distinguish 100% of protected and non-protected individuals with the CPS immunisation strategy have been identified. The average of the separation capability of all these type of identified classifiers is 95%.

Therefore, we have identified both linear and non-linear separators able to distinguish the 100% of protected and non-protected individuals immunized with the CPS immunisation according to the experimental validation carried out. Thus, a promising set of classifiers able to predict protection versus non-protection of individuals immunized with the CPS immunization have been identified.

Total RNA (including miRNA) was isolated from samples from a set of samples corresponding to a second CPS trial. Enough quantity and quality were obtained for all the samples. A list of 92 genes were selected for validation and all samples were analyzed by qPCR. Normalized Cts were incorporated in the mathematical model. Additionally, a miRNA profile of further CPS samples was carried out for the differentially expressed small and micro RNAs obtained from the analysis of the CPS experiment performed in WP1.

Results achieved
• Linear and non-linear separators able to distinguish 100% of protected and non-protected individuals immunised with the CPS strategy according to the experimental validation carried out.
• Separate samples from RTS,S immunised volunteers are being used for PCR validation of candidate biomarker signatures
• Additional blood and tissue cells collected and cryopreserved in NHP model for future investigation

Potential Impact:
Potential impact, dissemination and exploitation of results
We present two protective models of immunisation in this proposal for which to explore transcriptional immune signatures of protection. Such a finding is among the most sought after by malariologists and has enormous bearing on the future of malaria vaccines. Transcriptional signatures will reveal signalling pathways and activation of specific arms of the adaptive immune response, which will in turn provide insight to protective cellular mechanisms and generate new hypotheses that may be tested in experimental models. Most obviously, if a mechanistic correlate of protection is derived from this study, it will guide rational vaccine design by providing focus on the cellular pathways that must be stimulated for effective vaccination (e.g. TH1/TH2 T cell epitope selection; adjuvant selection).

Transcriptomics technologies are steadily improving and represent an attractive approach for tool development. However, transcriptome-wide screening is not practical for large-scale studies due to costs inherent in the technique. If the transcriptomics approach could be downsized to gene expression representative of specific biological responses, e.g. selected immune functions, then this would be highly valuable to researchers. Our results may guide development of an optimal tool, a tailored transcriptomics product for malaria vaccines, which may allow for cost-effective analysis of the genes defining the signature of protection. Additionally, our computational modelling allows for feeding the model of immune responses (transcriptomic) with other data (e.g. plasma and secreted cytokines). Together with a tailored transcriptomics tool, a product work package of in vitro assays may be designed as an improvement of the existing computational models to meet the needs of clinical research. An immediately realisable benefit of this new product would be easier data management, fewer artefacts and shorter analysis time due to fewer factors.

Impact on selection of vaccine candidates
Scientists need a methodology for selecting the most promising vaccine candidates for further evaluation while minimizing unfruitful investments in less promising or redundant approaches. If we identify combinatorial biomarkers of protection that may serve as a tool for predicting vaccine efficacy, we will also be solving the issue of how to prioritise the clinical development of current vaccines in the pipeline with a rational approach. Vaccine candidates (functioning with the same mechanism of protection) will have a standardised product work package with which to set benchmarks for rejection or further clinical development (e.g. vaccine efficacy greater than 60%, as determined by surrogate marker of protection). Furthermore, if our computational modelling approach proves that a signature of protection can be found, this will open the way for other vaccination models (e.g. blood stage vaccines) to perform similar modelling experiments using the Anaxomics platform to determine correlates of protection functioning under different mechanisms of protection. The end result will be accelerated selection and progress through clinical trials of viable vaccine candidates and a more cost-effective malaria vaccine product pipeline.

Impact on trial design
Use of a standardized surrogate marker of protection would allow for greater harmonisation of clinical trials. Throughout the 20+ year development of the RTS,S vaccine, trial designs have varied in their assessment of vaccine efficacy. Four types of follow-up have been used to evaluate efficacy: experimental challenge, active detection of infection, active case detection and passive case detection. In a trial design with a surrogate marker of protection, follow-up for evaluation of vaccine efficacy would be minimised, and the main measure of efficacy would be the results of the standardised assay(s). Patient follow-up would still be required, but the objective would be to evaluate safety by detecting severe adverse events, a much simpler follow-up. Comparability of results between trials in different epidemiological settings and with different adjuvants and antigens will be improved. Clinical trial harmonization will also have ethical and regulatory implications.

In a typical trial design using patient follow-up for analysis of protective efficacy, sample size is calculated based on expected efficacy of the vaccine, attack rate of disease expected in the control group (e.g. transmission intensity), and desired power (confidence levels) and significance level (e.g. alpha level 0.05). Sample sizes also very significantly by trial design, where active detection usually requires a smaller sample size than passive case detection. Under a trial design with surrogate marker of protection, the complexity of the trial is reduced, and the required sample size would, in most case, be significantly smaller.




Sample size calculation for a clinical trial of a malaria vaccine with estimated protective efficacy of 50% against clinical disease and a surrogate marker of protection with 100% predictive sensitivity/specificity (intention-to-treat):

n= (t^2 × p(1-p))/m^2 × C

n: required sample size; t: confidence level at 90% power (standard value 1.28); p: estimated prevalence of responders; m: significance level at 5% (standard value 0.05); C: contingency effect (10% to account for non-response or recording error, e.g. no blood sample collected)

Vaccine candidate: 181
Control vaccine: 181

In contrast, a similar trial of a 49% efficacious vaccine enrolled 894 participants (447 vaccine candidate, 447 control vaccine). In this case, a trial designed with a surrogate marker would have reduced the sample size significantly (here, 60%) and involved only 4 months of follow-up from baseline. Although, it must be considered that assessment of different endpoints such as severe malaria would require a higher n. It is impossible to calculate a fixed reduction or even an average reduction in the sample size of clinical trials through use of a surrogate marker design, given that each trial involves unique sample size calculations. However, the surrogate marker trial design clearly reduces the complexity of trials and most likely a sizeable reduction in participants, which in turn results in a significant cost reduction. Data management and monitoring would be simplified, as the current system involves monitoring teams that must handle all incoming hospital entries for identification of cases, and data managers must maintain a strictly kept demographic surveillance system and clinical database. Instead, routine QA/QC of the validated assays would be the primary requirement. Analysis of efficacy becomes a measure of the proportion of prevalence (of biomarker responders), rather than incidence rates between vaccine and control groups, which are affected by loss-to-follow-up (censoring) and can be further complicated by changes in transmission patterns over time. Ultimately, the greatest impact of our development of combinatorial biomarkers that predict protection will be realised in the re-design of clinical trials to be shorter and more cost-effective. For example, to develop optimal vaccination regimens by comparing the responder vs non-responder groups in small-scale trials.

Economic impact
This proposal aims to produce a set of combinatorial biomarkers that predict protection, which would be amenable as a surrogate marker. Given the 2-year scope of this project, the biomarker set (mostly transcriptomic, but possibly with translational outputs from other assays) will serve as a guideline for subsequent development of a tool for malaria vaccine evaluation. The results generated by the systems biology analysis in this project could be exploited for the design of an assay or product work package that may be used as a surrogate of protection for evaluation of malaria vaccine candidates and forgoing the need for lengthy clinical development. We speculate that there will be demand for the product from other vaccine developers in the malaria field, and that similar efforts will be pursued by developers of vaccines for other diseases.
Through this research, computational models newly implemented in vaccine systems biology have been developed using available datasets from the two unique malaria vaccine models towards meeting a clear biomedical need. Successful demonstration of this method would expand the Anaxomics portfolio of analytical tools to encompass the objectives of large canonical correlation studies (e.g. vaccine studies that induce complex immune responses), and would poise this platform to be adopted by other vaccine systems. Importantly, the usefulness and feasibility of the approach will be demonstrated and made available to the wider research and development community and help with the engagement of the Anaxomics with other.
The identification of a surrogate marker of immunity to malaria is a great step forward and sure to attract the attention of the private sector, particularly companies involved in vaccine and drug development. In the field of malaria vaccine development alone, the current vaccine candidate pipeline under development is filled.
An additional area of potential future growth will come from the basic research agenda, which will be interested in the mechanistic hypotheses resulting from these data. Such hypotheses can be taken up by academic investigators for testing in experimental models to further define and validate mechanisms of protection. Although this would be considered a bystander impact of the study, it would result in new studies, new grants, collaborations and ultimately investment into hypotheses derived from our initial results.

List of Websites:
Project public website
Project website address: www.sysmalvac.eu