BioCentury is providing this content for free given the urgent need for information about the coronavirus crisis. Further analysis can be found in our COVID-19 Resource Center. For more, sign up for our daily email.
1209 AZ

Product Development

AZ and Oxford paper spotlights the swerves on the road to pooled vaccine efficacy readout

A dive into AZ and Oxford’s efficacy analysis for AZD1222

A dive into AZ and Oxford’s efficacy analysis for AZD1222 reveals how the partners navigated data pooling in the face of protocol changes, and the questions now facing regulators.

Dec 11, 2020 | 11:14 PM GMT

The first peer-reviewed paper on pivotal data from a COVID-19 vaccine details the unplanned protocol shifts complicating the pooled efficacy analysis of AstraZeneca and Oxford’s AZD1222.

The Lancet study, published Tuesday, contains the efficacy and safety data for the adenoviral vector vaccine, which AstraZeneca plc (LSE:AZN; NASDAQ:AZN) and University of Oxford have submitted to regulators around the world.

It also provides details on two major protocol modifications: the decision to administer two vaccine doses instead of one, and the correction of quantification methods that led some participants to unintentionally receive a lower, but ultimately more effective, first dose.

These changes created unexpected differences across the four trials in the study, two of which the company combined into a pooled efficacy analysis; the other two trials, which didn’t have enough events to evaluate efficacy, were included in a pooled safety analysis.

In particular, the shifts generated wide variability in the time between the first and second vaccine doses, which was four weeks for some participants and more than 12 weeks for others. That variability in timing drove a further wedge between the subgroups and trials that were pooled together in the final analysis.

All the protocol changes, and the pooling strategy, were approved by regulators at multiple agencies before they went into effect, said Oxford Professor of Vaccinology Sarah Gilbert at a press briefing Tuesday.

Gilbert, a co-author on the study, led the team that designed AZD1222 at the university’s Jenner Institute. She added that the study only represents data collected through Nov. 4, and there is “a lot more to come.”

One question is whether regulators set to evaluate the interim data package will focus on the dosing regimen with more evidence but lower efficacy, and drill down on participant subgroups with similar time lags between doses.

Another is whether the company will seek to generate more data on the higher-performing lower prime dose. AZ EVP of biopharmaceuticals R&D Sir Mene Pangalos said the company had not decided on what its next trial would be. 

AZ and Oxford believe AZD1222’s efficacy, which was 70.4% overall across pivotal trials in the U.K. and Brazil, and 62.1% across all participants given two full doses, is sufficient for conditional approval, and will play an important role in the pandemic response despite higher performance by vaccines from Moderna Inc. (NASDAQ:MRNA) and partners Pfizer Inc. (NYSE:PFE) and BioNTech SE (NASDAQ:BNTX).

Among the subset of patients given a low first dose followed by a standard second dose, the efficacy of AZD1222 was 90%, closer to the competitors.

Pangalos highlighted the lack of severe COVID-19, hospitalization or COVID-19-related deaths among vaccinated participants at least 21 days after the first dose; all ten such incidents occurred in the control group. “The impact on healthcare systems, and the population at large, will be important.”

The partners have submitted data to MHRA, EMA, Health Canada, Brazil’s National Health Surveillance Agency, and other regulators around the world. Pangalos said that includes the U.S., but the assumption is that FDA will not consider authorizing AZD1222 until its U.S. trial reads out.

Dosing does it

Early data points from the partners’ first two trials prompted a series of protocol amendments over the summer. 

The switch from one dose to two was based on immunogenicity data from the Phase I/II COV001 study in the U.K, published in July. The results showed neutralizing antibodies in 100% of patients surveyed (9/9) from the trial’s small prime-boost arm, compared with 91% (32/35) of those studied who received a single dose.

AstraZeneca and Oxford announced the pivot when the paper went live, and protocol amendments for COV001, the U.K. Phase II/III study COV002 and the Brazilian Phase III study COV003 followed later that month. COV004, a Phase I/II South African study launched June 28, enrolled all participants to receive a double dose from the get-go.  

The dose strength issue was discovered earlier, but was not disclosed until the partners announced topline results in November. 

The issue came down to the methods used to quantify the adenoviral particles in vaccine formulations from three manufacturers: University of Oxford’s Clinical Biomanufacturing Facility (CBF); Advent Srl, a CMO unit of IRBM S.p.A. in Italy; and the U.K.-based viral vector CDMO Cobra Biologics AB, partnered with fill/finish CMO Symbiosis Pharmaceutical Services.

COV001 used a dosage of 5x10^10 viral particles as measured by spectroscopy, which was later shown to be 2.2 x10^10 viral particles as determined by qPCR. 

In consultation with MHRA, the COV002 sponsors planned to use the spectroscopy method to align with the first trial, but changed course after seeing the reactogenicity in trial subjects was lower than expected, and realizing that the formulation from Advent, which prioritized qPCR as a quantification method, had excess levels of the emulsifier polysorbate 80, which was throwing off spectroscopy readings.

With regulators’ approval, AZ and Oxford deemed that doses from Advent quantified via spectroscopy were a low, or “half” dose. Advent doses determined to contain 5 x 10^10 viral particles via qPCR, and all doses from Oxford’s CBF or Cobra/Symbiosis, were deemed a standard, or “full” dose.

The sponsors then developed a harmonized suite of assays to use for all standard doses going forward, with a target range of 3.5-6.5 x 10^10 viral particles per dose. That assay was used for all of the doses administered in the COV003 trial, and most of the doses administered in COV004; 44 participants received a low vaccine dose produced before adoption of the harmonized assays.

Study protocols in the paper’s appendix indicated that COV001 included both low doses and standard doses, but did not give a breakdown of how many participants received each type of dose, or whether there were changes in the study’s dose-quantification protocol. 

The changes in dosing strategy extended the lag between priming and booster shots because of time needed to ramp up manufacturing to support the two-dose strategy, and consult with regulators about the dose-quantification issue.

COV002 bore the brunt of the protocol switches, and participants who received a low priming dose were especially affected, with a median gap of 84 days, compared with 69 for COV002 participants who received two standard doses, 36 days for COV003 participants, and 28 days for COV005.

The variability in timing was also concentrated among participants 18-55, as older patients were not recruited into COV002 until after the two-dose regimen and harmonized quantification methods were already established.

An AstraZeneca spokesperson said these variations were in compliance with the trial protocols, which specified the interval between the first and second dose should be a minimum of four weeks apart. 

Pooling and slicing

Analyses combining and separating groups provide a mixed picture of AZD1222’s efficacy, and it will be up to regulators to determine what to prioritize and how to communicate the results on a label.

Andrew Pollard, chief investigator of the Oxford Vaccine Trial and a professor of paediatric infection and immunity at the university, said most of the data were generated to build a case for two full doses of the vaccine, and the 90% efficacy in the low priming dose subgroup from COV002 was likely to only be used as additional supporting evidence for the standard regimen.

He said the subgroup analyses in the paper indicate the low priming dose’s advantage in preventing symptomatic disease may be real, as it was maintained even after controlling for participants’ age and timing between doses. “It suggests the half dose is priming the immune system differently.”

Preliminary data suggesting the lower priming dose was also better at reducing asymptomatic cases in the COV002 study also bolsters the case, but it would be “premature to shout too loudly about it” given the low case numbers and wide confidence intervals on the results, Pollard said.

Among the three groups that have disclosed Phase III efficacy data for COVID-19 vaccines, AstraZeneca and Oxford are the only ones to have routinely screened a subset of participants for asymptomatic infection. 

The more immediate question is whether regulators will focus on the overall efficacy across all 131 cases pooled from COV002 and COV003 (70.4%), versus the pooled efficacy for the 98 cases that received two standard doses (62.1%), and how they will factor in the variability in timing.

Among the participants who received two standard doses of vaccine or control, 28 cases occurred in those with less than six weeks between doses, and 70 cases occurred in those with more than six weeks between doses. In these two subgroups, the vaccine had efficacies of 53.4% and 65.4%, respectively, with a wide confidence interval on the former, smaller group.

Pollard thinks the investigators are “not seeing a big difference in efficacy” based on the timing gap, and Pangalos thinks the variability in data comes with a silver lining.

“It actually enables you to immunize and protect more people if you have a slightly longer gap,” Pangalos said. “Having that flexibility in the dosing interval is actually quite a good thing, particularly during a pandemic.”

How to gain access

Continue reading with a
two-week free trial.