Experimental Infections in Humans, December 2003 – A Report for the European Malaria Vaccine Initiative

Alister Craig, Liverpool School of Tropical Medicine, Pembroke Place, Liverpool L3 5QA, UK

The next few years will see an increase in the number of malaria vaccines reaching phase III field trials but how should the candidates be chosen for this level of investigation. One of the tools available is the effect of interventions on experimental infections in human volunteers but should these be critical for the development pathway and how should they be conducted to minimise any risks at the same time as maximising the information produced? These were some of the questions addressed by a European Malaria Vaccine Initiative (EMVI)-sponsored workshop in The Netherlands.

Lessons learned from phase IIb trials (Chair: Blaise Genton, Basel/ Lausanne, Switzerland)
There are a range of parameters that need to be taken into consideration in a phase IIb trial, which can be broken down into;

1. Target Group – there was some discussion about what group of people should be targeted in terms of testing vaccines for malaria. In very young children (< 12 months) there could be problems with maternal protection and interactions with other vaccinations but in older children (> 36 months) acquired immunity may be a confounder, particularly in areas of high transmission. The consensus was that an age range of 12-24 months was optimal, although this might have a deleterious effect on acquisition of natural immunity (Judith Epstein, NMRC, USA).
2. Placebo – in some trials adjuvant alone has been used for the control group. This is useful as it controls for the effect of the adjuvant and provides safety data but there are questions as to whether it is ethical to give an adjuvant alone injection as part of a trial. One way to counter the ethical argument is to give another vaccine (e.g. rabies), although if the incidence of rabies is very low in the study area it is arguable whether this is any better than adjuvant alone. However this can cause difficulties with blinding samples and increases the cost of a study. As the combination of antigen and adjuvant is specific it is not clear if adjuvant alone would constitute an appropriate control (Tom Richie, NMRC, USA).
3. Measuring outcome – there are a number of ways of measuring malaria disease; time to infection; incidence; parasite density; first clinical episode; haemoglobin concentration; severe disease; mortality. The appropriate measure will be chosen partly in relation to the sample size (ranging from low to high in the list above). There are some problems with the non-disease measurements such as delay to infection as the link with clinical disease has not been proven (Vasee Moorthy, MVI, USA) and ITN trials that show a 50% reduction in mortality would not show any effect on parasite prevalence or time to infection (Adrian Hill, Oxford, UK).
4. Impact – in previous studies test populations have been cleared of any malaria infections by drug treatment. However this will cause alterations in the local environment (e.g. reduction in parasite densities) and adverse reactions to the drugs (albeit at low levels). In a study looking at the effect of Fansidar on a study looking at the Combination B vaccine (MSP1; MSP2; RESA), the power of the study in measuring clinical episodes was reduced by drug pre-treatment. One way to address this problem would be to use molecular techniques (genotyping) to measure new infections on a background of existing ones, which would remove the requirement for pre-treatment. However, this would need a high throughput system for genotypic analysis and would be difficult in erythrocytic stage vaccine trials where sequestration would be a confounder. Another option would be to carry out studies in areas of highly seasonal transmission but these might not be considered representative of the ‘normal’ situation.

Lessons learned from phase IIa experimental infections (Chairs: Tom Richie, NMRC, USA/ Robert Sauerwein, Nijmegen, The Netherlands)
Infection of human volunteers with P.falciparum has been carried out over several decades (Jeffrey et al, 1959), including therapeutic use as a treatment for neurosyphillis. The ability to carry out this type of work now is largely based on the relatively low morbidity and (in more recent times) lack of mortality seen in these studies. There are a number of elements that make up current experimental infection protocols;

1. Parasite strain - a number of different strains of P.falciparum have been used but current work has concentrated on 7G8 (Brazilian), 3D7 and its parent NF54 (probably West African). There are biological differences seen even within this small number of strains, particularly in sporozoite (spz) numbers in the salivary glands of infected mosquitoes (see below), but these could also extend to other processes including pathogenicity. Most work has been carried out using NF54 and 3D7 and there is good safety data for each strain.
2. Challenge – the main route for introduction of parasites into volunteers is by infected mosquitoes, although some work has been carried out using blood stage challenge (see below) (Frances Sanderson, Oxford, UK). There is some variation in the protocol followed by different groups but these fall mainly into two categories;
   a)
   Strain 3D7 (Oxford & NMRC) – five infected bites with at least 100 spz/ gland (although this is usually around 1,000 spz/ gland (grade 4)).
   b)
   Strain NF54 (Nijmegen) – five to seven infected bites with generally at least 10,000 spz / gland with an average of 150,000 spz/ gland.
   Both methods produce a 100% infection rate in volunteers but there are some significant differences in terms of the kinetics of the infection.
3. Parasite detection – the ‘gold standard’ for malaria diagnosis is the detection of parasites on thick and thin blood films. There is some variation in the exact protocol followed by different laboratories but the amount of blood screened is approximately the same. This is usually increased if there is a clinical suspicion of malaria. Infected erythrocytes can be detected at earlier times using quantitative PCR (18S RNA-based primers), which allows for monitoring of parasite levels prior to thick-film positive diagnosis and drug treatment. There can be considerable variation in the parasitological data (e.g. pre-patent period) but there tends to be a negative correlation between the number of infected bites and pre-patent period. This has serious implications for the length of time available for monitoring the effects of interventions in experimental infections.
4. Clinical aspects – in addition to the parasitological measurements there are a number of indicators of disease that are available in phase IIa experimental infection trials. Clearly these are not allowed to become severe and extensive clinical monitoring is part of all standard operating procedures (although there is some variation in how this is delivered). Most volunteers taking part in trials have some symptoms (mainly fever) and a few develop sufficiently "severe" symptoms to warrant admission to hospital, albeit for short periods of time and usually as a precautionary measure. In a review of 118 volunteers with experimental malaria infections only four had a temperature over 40oC and none suffered a parasite recrudescence (Church et al, 1997). This was supported by work presented at the meeting, indicating that the mosquito inoculation model for experimental infection is reliable and safe. Where clinical endpoints have been compared between control and intervention categories (fever; time to onset of symptoms) there were no significant differences between the two groups (Tom Richie, NMRC, USA).
   In rare cases volunteers develop symptoms of malaria before they become slide positive. Several approaches were suggested to address this issue including more frequent slides (or more extensive reading of slides) and PCR-based diagnosis but these can introduce bias into the study or not be readily available. In Oxford the standard procedure is when the managing physician diagnoses malaria, treatment is started and the time to patency is recorded as the time to treatment.
5. Modelling infections – the detailed longitudinal measurement of parasites in human volunteers has also facilitated estimations of the number of merozoites released from the liver and the erythrocytic multiplication rate. Different groups (Nijmegen and Oxford) used varying functions for their models which gave similar values for the number of merozoites per hepatocyte (20,000 – 30,000) and the multiplication rate (7.5 – 14 fold/ 48 hours).

Questions raised about the mosquito inoculation model

1. Is it necessary to use more than one strain of P.falciparum? Only three strains of parasites have been routinely used for volunteer studies and of these 3D7/NF54 (which are closely related) is the only one that is used routinely. Clearly it is important to measure responses to heterologous challenge as this is likely to be the situation in vivo, but this can be achieved by designing candidate vaccines based on non-3D7 strains and using 3D7 for the challenge. However, it is possible that parasite strains differ in their capacity to cause disease and so the concentration of research on a single strain may lead to misleading results. Against this is the unease that ethical committees and researchers might feel about introducing an unknown strain of P.falciparum into volunteers for possibly little gain in knowledge. For the moment it seems reasonable to continue using a small number of strains for these experiments, while remaining mindful of the limitations and continuing to monitor research into pathogenesis that might indicate any serious flaws in this path.
2. What level of challenge should be used? It is likely that there is a large difference in the number of sporozoites inoculated between the NMRC/Oxford protocol and that used in Nijmegen. This is most clearly seen in the shorter average time to patency seen in the former (11.5 days) versus the latter (9 days). As the onset of PCR positivity does not vary significantly between these two methods (at around day 6 after infection), the number of datapoints prior to drug eradication of the infection, when the volunteer becomes slide positive, is reduced using the Nijmegen protocol. This has been addressed by the Nijmegen group by delaying treatment for 48 hours after the volunteers become slide positive in a small group of people (n=5). Some increases in the parasite density and fever were observed but no serious adverse reactions were seen compared to volunteers treated as soon as they became slide positive. However there was a strong feeling of unease about this practice from other groups, who felt that this increased the risk of adverse effects too much and modified standard clinical practice which would be to offer treatment on the detection of malaria parasites in a thick film (regardless of whether the infection was experimental or not). All groups noted that symptoms became more severe after treatment, although this effect could be limited through the use of different drugs for parasite eradication (e.g by using Riamet (Arthemether + Lumefantrine)). As the reason for delaying treatment is to extend the data collection, it might be possible to create the same effect by reducing the level of the challenge.
   It is likely that both gland rates (103 and 104/105 spz per gland) fall within the range seen in natural infections and so it is difficult say which one represents more closely the level of infection seen in the field. A key question is whether the stringency of the challenge is a critical factor in human volunteer trails. The main outcome would seem to be whether the intervention makes a difference between treatment and control groups, which is not dependent upon whether the infection is more or less like natural ones. Given the difficulty in preparing pre-phase III trial information it could be argued that, at least initially, a lower stringency of challenge might be better with the possibility of a higher stringency follow-up.
   [NB. The use of a lower number of infected bites could be considered by the Nijmegen group]
3. What level of clinical monitoring should be offered to volunteers? There was no difference between groups in the basic premise that volunteers should receive a high level of clinical care. The way in which this was administered varied, mainly due to geographical constraints, such that in smaller cities clinical examination twice a day (blood film plus sample for PCR) was considered sufficient but in larger areas volunteers were gathered in hotels overnight (from day 4 post-infection) where direct access to clinical support could be offered. Treatment to eradicate parasites was given usually when thick blood films showed the presence of infected erythrocytes or when the managing physician determined that the clinical symptoms were sufficient to diagnose malaria. Withholding treatment for 48 hours after parasites were detected on thick films did not result in any severe adverse reactions but was generally considered not to be acceptable. In rare cases, support beyond the trial period was needed (and given).
4. Should phase IIa experimental infections form part of the critical development pathway for malaria vaccines? There was no doubt that a positive result in experimental infection trials would strengthen the case for a vaccine candidate proceeding into phase IIb/ III clinical trials. What was harder to derive was whether a negative result would preclude a vaccine from further study. In many ways this mirrors the arguments for non-human primate studies, which almost certainly have a place in vaccine trails but have not always been used to derive GO/ NO GO criteria. While in non-human primates there might be an argument that the mechanism of immunity is not identical to humans, this could not be argued for human volunteer trials. Thus it would seem to be reasonable that where an experimental infection study has been carried out in humans, it should form part of a GO/ NO GO decision. However, given the difficulties in carrying out human volunteer trials (the technical difficulties and resource implications for this type of work are very high) it is debatable whether these should be critical to the development of a malaria vaccine.
5. Can the mosquito inoculation model be used to test blood stage vaccines? In theory the mosquito inoculation model is well-suited to testing all vaccines, including blood stage targets, as it mimics natural infections (e.g. cytokine release during the liver stage may influence the blood stage). However, in practice the window of action for interventions on blood stages using this model may be too short. An alternative model using blood stage challenge was presented (Frances Sanderson, Oxford, UK) using P.falciparum infected blood stages raised in screened blood (HIV; Hep B/C; Syphilis; Ross River Virus). 1,200 infected erythrocytes were injected into five volunteers who became thick film positive on days 7-9 post-infection. This approach offers a way to introduce controlled numbers of infected cells, thereby extending the pre-patent period. More work is required to ascertain the reproducibility of this assay and to study the immune effectors used in this type of trial.

Summary
Overall there was strong support for the use of experimental infections in human volunteers in the development pathway of malaria vaccines, particularly for pre-erythrocytic stage candidates. There was some discussion about the details of the mosquito inoculation model, particularly in terms of stringency of the model.

Acknowledgements
I would like to thank Robert Sauerwein for his comments on this report, the organisers of this workshop, particularly Barbara Schimmer and Loes Borst, and all the participants for their lively and energetic contributions to the meeting, on which this review is based.

List of participants

Pierre Druilhe, Institut Pasteur, Paris France
Judith Epstein, NMRC, Rockville, USA
Valerie D’Acremont, University of Lausanne, Lausanne, Switzerland
S. de Vlas Erasmus, Medical Center, Rotterdam, The Netherlands
Blaise Genton, University of Lausanne, Lausanne, Switzerland
Rob Hermsen, UMC, Nijmegen, The Netherlands
Adrian Hill, University of Oxford, Oxford, UK
Soren Jepsen, EMVI, Copenhagen, Denmark
Odile Leroy, EMVI, Paris, France
Shirley Longacre, Institut Pasteur, Paris France
Vasee Moorthy, MVI, Rockville, USA
Z. Reed, WHO, Geneva, Switzerland
Tom Richie, NMRC, Rockville, USA
Christine Sadorge, Institut Pasteur, Paris France
Frances Sanderson, University of Oxford, Oxford, UK
Robert Sauerwein, UMC, Nijmegen, The Netherlands
Barbara Schimmer, UMC, Nijmegen, The Netherlands
Martin Steiner, Apovia AG, Martinsried, Germany
A. van der Ven, UMC, Nijmegen, The Netherlands
D. Verhage, UMC, Nijmegen, The Netherlands

References

Church, LWP, Le, TP, Bryan, JP, Gordon, DM, Edelman, R, Fries, L, Davis, JR, Herrington, DA, Clyde, DF, Shmuklarsky, MJ, Schneider, I, McGovern, TW, Chulay, JD, Ballou, WR and Hoffman, SL (1997) Clinical Manifestations of Plasmodium falciparum Malaria Experimentally induced by Mosquito Challenge, J. Infect. Dis. 175: 915-920

Jeffrey, GM, Young MD, Burgess, RW and Eyles, DE (1959) Early Activity in Sporozoite-Induced Plasmodium falciparum infections. Ann. Trop. Med. Parasitol. 53: 51-58

Shute, PG and Maryon, ME (1966) Laboratory Technique for the Study of Malaria. p59. J & A Churchill, London