AstraZeneca’s Phase I/II Covid-19 vaccine trial unlikely to report convincing protection data

Although the AstraZeneca Phase I/II AZD1222 Covid-19 vaccine trial has a co-primary efficacy endpoint investigating the vaccine’s potential to protect against symptomatic disease, the 1,000-volunteer trial is not powered enough to demonstrate a clinically significant signal of the vaccine’s efficacy.

Phase I/II of the AstraZeneca trial only requires 30 Covid-19 cases to detect vaccine efficacy, which is too small and would only validate the ongoing, larger trials. Any early stage trial is designed to only investigate immunogenicity, not protection from disease. In a 15 July media interview, AZD1222 principal investigator Sarah Gilbert said the vaccine’s efficacy could be known by September.

Aside from its Phase I/II trial, AstraZeneca has indicated pooling data from the various large Covid-19 vaccine trials in the UK, Brazil and South Africa could offer the possibility for quicker protection data readout. However, combining results from these trials as they have different designs is sceptical. Because these trials have various protocol differences, such as dosing and dosing schedules, merging the data would impact the value of the reported outcomes.

While the UK-based Phase I/II (NCT04324606) has reported safety and immunogenicity data, results from the coprimary efficacy endpoint of protection from symptomatic disease, which has a six-month timeframe, are yet to be revealed. Volunteers were dosed in April and May. AZD1222 is already in three ongoing large trials in the aforementioned countries, with a US-based trial yet to recruit patients.

In a 21 May media release, AstraZeneca has indicated it has secured total manufacturing capacity for one billion AZD1222 doses, with the first deliveries expected in September. The University of Oxford’s Jenner Institute initially developed the vaccine before partnering with AstraZeneca on 30 April.

When asked to comment, an AstraZeneca spokesperson referred to this news service to the University of Oxford. The university declined an interview request and did not respond to a request for comment.

AstraZeneca Phase I / II Covid-19 vaccine trial underpowered to demonstrate protection

On 15 August, safety and immunogenicity data from the Phase I/II AZD1222 trial was published in The Lancet. In its supplementary appendix, which was included as part of the original submission for trial approval, the coprimary endpoint vaccine efficacy measure will be calculated as (1 – the relative risk of symptomatic infection) x 100%.

An analysis of the coprimary endpoint will be conducted when at least 30 Covid-19 cases have occurred. With 30 cases, the trial will have 80% power to detect 70% vaccine efficacy, assuming an event rate of 23 cases in the control arm and seven cases in the AZD1222 arm.

Having such Covid-19 cases in each arm would give the data a p-value of around 0.004, which would be statistically significant, noted Lawrence Moulton, PhD, professor, Department of International Health, Johns Hopkins University, Baltimore, US. Phase I/II volunteers received a single shot of the vaccine, except for ten participants who received two doses. Volunteers in the control arm received the meningococcal vaccine MenACWY.

However, the appendix notes the Phase I / II is only designed to detect an efficacy signal, meaning the lower bound of vaccine efficacy of the trial is 0%, Moulton added. According to a June 2020 FDA document detailing industry guidance on Covid-19 vaccine development, the primary efficacy endpoint for a placebo-controlled efficacy trial should be at least 50%, and the statistical success criterion should be the lower bound point estimate of 30% or more.

Any data from the Phase I/II would lack clinical significance and worthiness of authorisation from regulatory authorities, as such data would only be enough to validate continued investigation, noted Ira Longini, PhD, co-director, Center for Statistics and Quantitative Infectious Diseases, Emerging Pathogens Institute, Gainesville, Florida, US.

Even if there are no cases in the vaccine arm in Phase I/II’s first seven total cases, it would only justify larger trials, added Peter Gilbert, PhD, biostatistician, Fred Hutchinson Cancer Research Center, Seattle, Washington.

With various large-scale AZD1222 trials already ongoing, questions remain about what would happen to the AstraZeneca Covid-19 vaccine development plans if the Phase I/II trial did not report statistically significant results, noted Susan Ellenberg, PhD, interim chair, Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, US.

Out of the 30 Covid-19 cases, if 17 were in the comparator arm and 13 in the vaccine arm, this would mean the vaccine would offer protection by chance alone, she explained. However, having a low success bar in Phase I/II is not unprecedented, with some Phase I/II vaccine trials reporting of a p-value as high as 0.15 still moving on to Phase III, Moulton said.

In contrast to other early-phase Covid-19 vaccine trials, AZD1222’s Phase I/II actually has an unusual coprimary endpoint looking into the number of virological confirmed, polymerase chain reaction (PCR)-positive, symptomatic cases, Moulton said. Phase I/II trials are not typically powered enough to demonstrate protection and are only used to show immunogenicity and early safety signals, Longini explained.

The reason for having early-phase primary endpoints looking into immunogenicity is that results would be collected more quickly, at one month after vaccination; a protection primary endpoint, such as with AZD1222, would take longer, Moulton noted.

Protection data at Phase I/II stage takes longer because those volunteers are typically at lower risk of infection compared with Phase III volunteers, which slows the Phase I/II in reaching its Covid-19 case number milestone, he explained.

In Phase III, volunteers are intentionally selected to be at high risk of contracting the virus to reach the trial’s case number target, he added. In Phase III, Brazil-based trial, volunteers at high risk of infection, including healthcare workers, are recruited.

Large trials challenging to forecast

Although the Phase I/II may have a low protection success bar, this does not mean its larger trials would also be the same, said Natalie Dean, PhD, assistant professor, Department of Biostatistics, College of Public Health & Health Professions, University of Florida, US. The statistical rationale behind AZD1222’s large trials has not been made public. If the Phase I/II reports of a positive coprimary protection endpoint, it could bode well in the larger trials, Moulton said.

Still, the Phase I/II is too small a trial to forecast how the larger trials would perform, Dean said. Although AZD1222 has Phase I/II immunogenicity data, it is challenging to identify what immunogenicity data is needed to predict protection, this news service reported on 16 April.

A Phase I/II AZD1222 secondary endpoint is looking at quantifying antibodies against SARS-CoV-2 spike protein, with a six-month timeframe. With an antibody endpoint, it could provide some insight into AZD1222’s protection against infection if 10% of volunteers in the comparator arm are infected, regardless of disease severity, Moulton said.

In the Lancet article, neutralising antibody responses were detected in 32 out of 35 participants after a single AZD1222 dose, and in all nine participants receiving a booster dose.

However, AZD1222’s protection against symptomatic infection, the Phase I/II coprimary endpoint, is more clinically significant than protection from infection alone, Longini said. There is value in measuring infection rates as an insight into the prevention of viral spread, he noted. However, such an analysis would be conducted post approval as transmission rates would need to be measured, and they require a different set of investigative tools, he explained.

Worrisome pooling data from disparate trials 

The aforementioned Lancet supplementary appendix also states both UK-based Phase I/II and Phase II/III trials are expected to be running concurrently, so efficacy data from both trials will be combined in a prospective meta-analysis for a more precise estimation of efficacy and safety.

It will likely be difficult to interpret such data as earlier-phase trials tend to skew towards younger, healthier participants, Moulton and Gilbert added. Both the Phase I/II and II/III trials have arms recruiting volunteers ages 18–55 years, but Phase I/II participants have a median age of 35 years and an interquartile range of 28–44 years. Still, a meta-analysis for safety is typical for a cumulative safety profile, Moulton explained.

During AstraZeneca’s 30 July investor call on its 1H20 results, the company indicated the pooling of its large trials in the UK, Brazil and South Africa for regulatory filing is possible due to their similar primary endpoints. The Phase II/III (NCT04400838) in the UK is recruiting around 12,300 participants, the Phase III in Brazil is enrolling 5,000 volunteers, and the Phase I/II South Africa-based trial (NCT04444674) is recruiting 2,000 participants.

While the Brazil trial is not listed on ClinicalTrials.gov, the primary efficacy endpoint for the UK-based trial is the number of virologically confirmed, PCR-positive, symptomatic cases of Covid-19 with a six-month timeframe, and the South Africa trial has a 12-month timeframe investigating the reduction of severe and nonsevere Covid-19 disease compared with placebo.

To allow for the pooling of data, the trials should have a unified framework, which does not seem to be the case with AZD1222 based on the large trials’ ClinicalTrials.gov pages, Dean said. Combining data from different trials, even if comparable, is already challenging as there are obstacles like the need for consistency on how data is collected, Ellenberg said. Trials usually have different monitoring plans, which makes it challenging to pool results, Dean added.

One notable difference between the large trials, based on their ClinicalTrials.gov entries, are the various doses used. In the UK-based Phase II/III, the primer injection uses the 5×10^10 viral particle (vp) dose, with the booster dose of 2.2×10^10vp administered four-to-six weeks later. In the South Africa trial, both primer and booster use the 5×10^10vp dose, with the booster administered four weeks later.

In the Lancet article, neutralising antibody titers were strongest after a booster, suggesting a second dose is ideal for stronger protection. Both UK and South African trials have a single dose and primer-boost vaccine schedules, while volunteers in Brazil only receive a single shot.

Pooling data from trials with different doses could dilute results to the point it may be easier to assume a higher dose is the most efficacious approach, Ellenberg said. The US-based, 30,000-volunteer Phase III (NCT04516746) is yet to start recruiting, but its entry shows it is using the South Africa dosing schedule, with a nominal difference of ±1.5×10^10vp between individual doses.

For more insight and data, visit GlobalData's Pharmaceuticals Intelligence Centre.