A Dose (or more) of Positivity - Reading into AstraZeneca's Interim Analysis Results

It’s three from three as far as positive outcomes from COVID vaccine trials are concerned but Monday’s announcement from AstraZeneca and Oxford University, at a first glance, may not seem to be as exciting as those from Pfizer and BioNTech, and Moderna. Furthermore, the figures are a bit of a head scratcher, so let’s look at them in more detail.

Today’s results from AstraZeneca and Oxford University state that the combined Phase III interim analysis from their COV002 and COV003 studies (based in the UK and Brazil respectively) included 131 COVID-19 cases and that the vaccine was found to be 70.4% effective overall, but that vaccine efficacy in two dosing subgroups was 90% in one and 62% in the other. Unfortunately, the complex design of this trial, in comparison to the Pfizer and BioNTech, and Moderna studies means that we are unable to examine these numbers in more detail and estimate how many COVID-19 cases could have been in each group. According to the trial protocol for the UK study (COV002), the AstraZeneca and Oxford University study, participants were recruited in different age groups: 18-55, 56-69, and 70+. Those in the 18-55 age group were randomised 1:1 to either ChAdOx1 nCOV19 vaccine or MenACWY (control). But for those in the 56-69 age group, randomisation was 3:1 candidate versus control, and for those aged 70+, randomisation was 5:1.

The deviation from 1:1 randomisation is interesting and hasn’t been seen with the other studies published to date. But allocating more participants to the candidate vaccine may have advantages in the collection of safety data, for example. A larger sample size for those taking the ChAdOx1 nCOV19 vaccine gives more power to detect adverse events.

What is interesting from Monday’s announcement are the results surrounding the different dosing regimens. We are told that one dosing regimen (with a sample size of 2,741 participants) showed VE of 90% when the ChAdOx1 nCOV19 vaccine was given as a half dose, followed by a full dose at least one month apart, and another dosing regimen (with a sample size of 8,895) showed 62% VE when given as two full doses at least one month apart. The estimated VE for the half dose gave a higher VE than that estimated for the full dose. This seems counterintuitive and unexpected. But the sample size for the half dose was smaller than for the full dose and so the point estimate of 90% VE for the half dose will be subject to more uncertainty than for the full dose.

Furthermore, we are told that the reason for the half dose being considered was actually an accident. Mene Pangalos, head of AstraZeneca’s non-oncology research and development told Reuters that the plan was for trials participants to receive two full doses, but that researchers were puzzled when they noticed fewer side effects than expected. Pangalos says “So we went back and checked ... and we found out that they had underpredicted the dose of the vaccine by half.” It will be interesting to see the full results of this accident when they are published, but questions that immediately come to my mind are:

  1. This accident only happened in the UK, and quite possibly only at one site. It may even have only happened within one age grouping. We must then ask ourselves how generalisable are the results to the worldwide population? It is likely that further investigation into the efficacy of this dosing regimen will be needed on a wider scale.
  2. It is perfectly acceptable to make changes to the protocol prior to database lock, so the protocol could have been updated to include this additional analysis. However, according to version 14 of the protocol, dated 9th November 2020, the primary analysis was set to be the efficacy of two doses of vaccine (across both half and full dose), with secondary analyses being the efficacy of at least one full dose and efficacy of two full dose of vaccine. Efficacy of half dose with a full dose booster was not considered as a secondary analysis in the protocol and so could be an ad-hoc analysis post database lock.

There have been numerous comparisons made between the AstraZeneca results announced today and the previous announcements from Pfizer and BioNTech, and Moderna, but there are several reasons why that isn’t a sensible thing to do. Direct comparisons of results from different trials should generally be done with caution due to differences in study population, endpoints, and regions. If one wishes to carry out a comparison of the vaccine efficacy of multiple candidates, this should be done in a head-to-head clinical trial. Otherwise you are not comparing apples with apples. For example, in the Moderna study to be considered a case of COVID-19 for the evaluation of the primary VE endpoint, the following criteria needed to be met:

  • The participant must have experienced at least TWO of the following systemic symptoms: Fever (≥ 38ºC), chills, myalgia, headache, sore throat, new olfactory and taste disorder(s), OR
  • The participant must have experienced at least ONE of the following respiratory signs/symptoms: cough, shortness of breath or difficulty breathing, OR clinical or radiographical evidence of pneumonia; AND
  • The participant must have at least one NP swab, nasal swab, or saliva sample (or respiratory sample, if hospitalized) positive for SARS-CoV-2 by RT-PCR.

For Monday's AstraZeneca and Oxford University results, a COVID-19 case was defined as meeting the following criteria:

  • A participant with a positive PCR swab, AND
  • At least one of the following symptoms: cough, fever ≥ 37.8, shortness of breath, anosmia, or ageusia.

So the way that COVID-19 cases were defined were different in each of the studies. Furthermore, both the Pfizer and BioNTech, and Moderna studies were placebo-controlled, whereas the AstraZeneca study used the meningococcal vaccine MenACWY as the study comparator. Finally, we also know that the different results correspond to different populations. Monday’s results from AstraZeneca and Oxford University are based only on participants based in the UK and Brazil, with clinical trials being conducted in the US, Japan, Russia, South Africa, Kenya and Latin America, and planned trials in other European and Asian countries. The Pfizer and BioNTech results, for example, are global results, meaning that the two estimated VEs are not directly comparable.

Monday’s announcement from AstraZeneca and Oxford University is a positive one and although 2020 will undoubtably be remembered by the number of lives lost, the hours spent in lockdown, and the number of zoom calls with our colleagues, let’s never forget that 2020 should also be remembered as the year that science provided us with three working vaccines developed at breakneck speed. These certainly won’t be the final results that we’ll get from AstraZeneca. There will be more numbers coming through from the other studies taking place worldwide. Furthermore, we are given no information today about the estimated VE across age groups and ethnicities. We are told that there were no severe COVID-19 cases or hospitalisations in the vaccine group, be we aren’t told how many there were in the control group (if any). This study also carried out monitoring for asymptomatic infection and we are yet to see these results although we are told that “Early indication that vaccine could reduce virus transmission from an observed reduction in asymptomatic infections”. Crucially, this vaccine has a huge advantage over others in the way that it is stored. This vaccine will be easily administered in existing healthcare systems, stored at fridge temperature and distributed using existing logistics. Vaccine efficacy is just one part of the puzzle. Here we have a vaccine that could demonstrate vaccine effectiveness of around 90% and be easy to store and distribute. That is definitely something to celebrate.

Three from three…I’d say that’s pretty incredible!