Suggested Design of Trials for Testing Malaria Vaccines in Nonimmune Adults Visiting Endemic Areas
A total of three critical phase 3 trials of efficacy is envisioned. All three trials would be randomized (at the level of the individual subject), controlled, and double blind (or blinded observer). In the two initial small trials, the subjects, under close clinical supervision, would not take concomitant chemoprophylaxis. In the third study, the participants would be given chemoprophylaxis, but it is assumed that at least 10 percent of them would not take it. Clinical supervision would still be provided. In this way the study would be analogous ethically to experimental challenge studies where subjects who do not take chemoprophylaxis are exposed to the bites of five Anopheles mosquitoes infected with P. falciparum.
TRIALS IN THE ABSENCE OF CHEMOPROPHYLAXIS
In the initial two small phase 3 efficacy studies in Western Kenya (or perhaps Ghana or Indonesia), the subjects would be entirely dependent on the accompanying medical staff to provide prompt diagnosis of malaria, to initiate optimal specific therapy, and to maintain follow-up to avoid complications. During these two relatively small initial efficacy trials, the opportunity would be taken to collect sera and peripheral blood mononuclear cells from the subjects at baseline and at various time points thereafter to perform measurements of serum antibodies and cell-mediated immune responses. If the vaccine proves efficacious, the hope would be to identify immunologic correlates of protection.
To have 90 percent power to detect a statistically significant differ-
ence (alpha = 0.025, single tail) in the attack rate for clinical malaria in vaccinees versus controls (based on estimated 60 percent vaccine efficacy and a lower limit of 30 percent for the 95 percent confidence interval [CI]), the trial would have to be large enough to allow detection of a total of 160 confirmed P. falciparum clinical malaria cases.
Some assumptions for design of the trials without chemoprophylaxis include the following:
At least 70 percent of the U.S. control subjects will develop clinical malaria during approximately 5 months of stay in peak transmission season. (It is recognized that this is likely a conservative estimation as the attack rate in controls is more likely to approach 100 percent).
Because of the remoteness of the geographic location, the duration of local exposure (approximately 5–6 months) and the other demands of participating in an intensive, complex vaccine trial, a dropout rate (loss to follow-up) of up to 18 percent must be expected.
A total of 164 analyzable subjects per group is needed to have 90 percent power to detect a significant difference (alpha = 0.025, single tail) if the expected attack rate in controls is 70 percent and expected vaccine efficacy is 60 percent (with a lower limit of 30 percent for the 95 percent CI). If 200 subjects are randomly allocated to the malaria vaccine group and 200 to the control group, with 18 percent loss to follow-up, at the end of the study there will remain approximately 164 vaccine and 164 control subjects available for analysis. At the expected attack rate, this would yield about 115 confirmed P. falciparum cases among the controls and about 46 cases among the vaccinees (60 percent proportionate reduction); the 161 cases in this scenario would provide the total of 160 cases needed to address the primary aim. With these results as an example, the 95 percent CI around the 60 percent point estimate of vaccine efficacy would be 43 percent lower limit and 72 percent upper limit.
If this first phase 3 efficacy trial in subjects not under cover of chemoprophylaxis is successful, the committee proposed that a corroborating trial of identical design be carried out one season later. This trial would provide a second opportunity to collect clinical specimens in the search for immunologic correlates of protection.
TRIALS IN THE PRESENCE OF CHEMOPROPHYLAXIS
If the corroborating phase 3 trial not under chemoprophylaxis also yields positive results, it would be appropriate for the Military Infectious Diseases Research Program (MIDRP) Malaria Vaccine Program to undertake a much larger phase 3 trial with 10 times as many subjects in the
vaccine and control groups, all of whom would be issued standard military chemoprophylaxis. Although subjects would be recommended to take chemoprophylaxis, there would be no systematic direct supervision of subjects taking their daily medication. Rather this would be left to the discretion of the individual subject, recognizing that in real-life conditions, a variable proportion of military personnel deployed to sites of known malaria risk do not take chemoprophylaxis in a reliable way. Accordingly, in a conservative assumption, 10 percent of the study subjects would fail to take chemoprophylaxis for sufficiently extended periods so that these subjects would be equivalent in risk to the nonprophylaxed subjects of the preceding two efficacy trials.
Thus, if 2,000 enrolled subjects were randomly allocated to receive the maturing candidate vaccine and 2,000 others to the control group, by the end of the study, despite some expected dropouts and loss to follow-up, approximately 1,640 analyzable subjects would be available in each group. Of these, because of random allocation, one would expect about 164 “nonchemoprophylaxed” subjects to be available for analysis in each group. Among the 1,640 analyzable control subjects, one would expect to detect around 115 cases of P. falciparum malaria (70 percent attack rate among the 164 controls who did not adhere strictly to chemoprophylaxis). One would also expect to detect 46 cases of P. falciparum malaria in the vaccine recipients (60 percent proportionate reduction); this constitutes a total of 161 cases between the two groups. The limits of the 95 percent CI around the 60 percent point estimate of vaccine efficacy, as in the previous example, would be 43 percent (lower limit) and 72 percent (upper limit) around the point estimate of efficacy.