It’s very complicated, so no wonder there has been a lot of confusion. Errors, information gaps, and mistaken assumptions were common in the rush of coverage of the press release of interim phase 3 results for the Covid-19 vaccine developed by Oxford University, its spinout company Vaccitech, and AstraZeneca. This post summarizes information I have gathered to keep track of clinical research progress for this vaccine. It provides a bibliographic base for my post at WIRED about some aspects of the interim phase 3 analysis. (You can see my other WIRED posts here.)

If you are looking for specific documents, you can also browse through my collection of studies and trial-related records on Covid-19 vaccines published by their developers to see key records for this trial. (There’s an explanation of this collection and how to use it at the end of this post.)




  • First human trial of this vaccine began: a phase 1/2 trial in the UK (called COV001), with a comparison group receiving meningococcal vaccine.


  • Soon after the phase 1/2 trial began, study arms were added with 6-hourly doses of acetaminophen (paracetamol) for 24 hours, in an attempt to reduce the rate of adverse events.




  • A preprint of a preclinical research for the vaccine was released, receiving criticism for over-optimistic interpretation. It has since been published in Nature. If you’re interested in preclinical research, there are 4 records for these studies on this vaccine.


  • The Oxford vaccine became the first non-Chinese Covid-19 vaccine to reach phase 3 trial stage, with a phase 2/3 trial in the UK. It was originally registered with 14 arms, aiming for 10,260 participants. And again, the comparison group was meningococcal vaccine.







  • This month, they started a phase 1/2 trial for 2,000 participants in South Africa, to see how the vaccine worked in people who were HIV positive compared to people who were HIV negative. It wasn’t very standardized: for example, follow-up and taking blood samples varied, and there were arms where people getting one and two doses were mixed together.


  • BNT/Pfizer announced if all went well with earlier tests, they would go into a 30,000-person phase 3 trial in the US in late July. And they did: it started on July 27, the same day as Moderna’s 30,000-person phase 3 trial in the US.


  • The results of the phase 1/2 trial of the Oxford/AstraZeneca vaccine in the UK were published, and the protocol for the trial was published with it in an appendix. The adverse effects were indeed high: I wrote it about this at WIRED. While promoting these findings, the developers said they, too, were planning a US trial with 30,000 people.





  • An update on recruitment progress on August 14: 8,000 people in the phase 2/3 trial in the UK; 3,032 in phase 3 trial Brazil; 914 in phase 1/2 trial in South Africa.


  • A phase 2/3 trial started in India for 1,600 participants. It compared the Oxford/AstraZeneca vaccine with saline placebo, and a locally-produced version of the Oxford/AstraZeneca vaccine.


  • Further changes were made to the trial register entry for the phase 2/3 trial in the UK, some of which related to exploring whether an injection was inducing immunity to the harmless virus vector component of the vaccine.


  • A phase 3 trial in the US for 30,000 participants was registered, with a saline placebo arm.






  • On September 6, the trials for this vaccine were put on hold around the globe, while a serious adverse event in the UK trial was investigated. It was the second time there had been a hold to investigate a serious adverse event, a standard process in clinical trials. I detail the steps in the early weeks here. Although details provided publicly are sketchy, both were apparently neurological events, one of which was assessed as unrelated to the vaccine; the other was not deemed related to the vaccine but was not cleared of a relationship either. All the trials except the US one resumed fairly soon.



  • They applied to the Brazilian authorities to expand the trial to 10,000 people.




  • Details of AstraZeneca’s contract with the Brazilian government were reported: the government is to pay US$300 million, even if the vaccine is never approved. (These contract arrangements are usually not made public.) 


  • On October 12, the trial register entry for the UK phase 2/3 trial was updated again, expanding the trial to 28 arms and 12,390 participants.


  • The death of a participant in the Brazilian trial was reported. The trial would later be halted for a short period of time. Although there was no official statement from the company or drug regulator, there were reports that the person was in the placebo group and died by suicide.


  • On October 23, the FDA cleared the US trial to resume, without providing details of the serious adverse event that had put the trial on hold. At the end of the month, the US trial was expanded to 40,051 anticipated participants, including from sites in Latin America.


  • A trial of 400 people to test immune responses in the local population was announced for Kenya. (I haven’t identified a trial register entry for it, and haven’t tried local registries.)





  • A phase 3 trial with 300 people was announced for Italy. (I haven’t identified a trial register entry for it, and haven’t tried local registries.)


  • Sometime in the early-mid part of the month, the Brazilian trial was reportedly fully recruited with over 10,000 healthcare workers, but was seeking additional participants over the ages of 55 and 65.


  • On November 9, BNT/Pfizer became the first vaccine to release an interim analysis for a large phase 3 trial, indicating that the vaccine efficacy rate was over 90%. And they released an updated protocol for the trial, which had grown to over 43,000 participants. (I wrote about the press release’s results at WIRED.)


  • On November 11, the Serum Institute of India reported that the now phase 3 trial there for 1,600 people was fully recruited, and they intended to apply for emergency use authorization soon. They had manufactured 40 million doses in anticipation.


  • November 16, and Moderna announced its interim analysis for its 30,000-person trial, with 95% vaccine efficacy.


  • BNT/Pfizer announced on November 18 that they had reached their final minimum number of events (170 people with symptomatic Covid-19), with 95% vaccine efficacy after 2 doses, including 94% efficacy in people over the age of 65.


  • And on the same day, November 18, the Oxford/AstraZeneca vaccine’s phase 2 results from their phase 2/3 trial in the UK were published in The Lancet. The media hype for this study was on signs of immunogenicity in older people. But I thought the really big news here was what was in the appendix. Just as they had with the phase 1/2 previously, they published the trial protocol with the results. And there was an unexpected plot twist: instead of individual interim analysis plans for each of their very different trials, they were proposing a meta-analysis of 2-dose groups from 4 extremely different independent trials: the phase 1/2 trials in the UK and South Africa, the phase 2/3 trial in the UK, and the phase 3 trial in Brazil. I explained it in a tweet thread. And as I stress in an explainer post about understanding data in meta-analyses, a key issue is not combining trials that aren’t similar enough. Given the radical concept here – using meta-analysis for efficacy without a major single phase 3 trial – the protocol is very light on details about this proposal. And they didn’t specify what number of events (symptomatic Covid-19 illness) would trigger this combined analysis. On November 19, in an interview, they said it would be 53.


  • On November 20, there was a report that Anvisa, the Brazilian FDA equivalent, had publicly registered a change in the statistical analysis plan for the Oxford/AstraZeneca phase 3 trial in Brazil. There were no details of what it was, according to the report, but for it to be registered in that way meant it was something important.


  • On November 23, Oxford/AstraZeneca released an interim analysis that pooled the results of some subgroups of just the UK phase 2/3 trial and the Brazilian phase 3 trial. I discuss it in some detail at WIRED. It doesn’t seem likely to me that the timing of the change lodged by Anvisa and this analysis is a coincidence.


Where does this leave us? I discuss issues and red flags in detail at WIRED, but a few major issues didn’t fit there. The press release doesn’t provide the most basic level of detail to enable us to understand what they did: we don’t know how the percentages were derived, exactly what the comparisons were, or what the level of uncertainty (confidence intervals) is around the numbers they reported. We don’t know if they had enough severe Covid-19 events to draw meaningful conclusions about that critical issue. [Update: There apparently were, according to an interview with the head of Operation Warp Speed.]

Given the enormous difference in when second injections were given – in the Brazilian trial it could be up to 3 months – we really need to see what happened to the people in that interval. The primary analyses for efficacy are done after the second injection – but to get an idea of likely effectiveness, we also need to know what happened in between shots, and how many people didn’t get that second shot at all. That’s important for any two-shot vaccine, but it’s particularly critical for one with a high rate of adverse events.

Bottom line? When you consider the press releases from BNT/Pfizer and Moderna alongside the detail they report in their protocols and for their very large coordinated trials generally, it’s easy to be confident those vaccines work very well – even though there’s still a lot to know.

With the Oxford/AstraZeneca vaccine, though, the combination of what we know about their trials, the twists and turns of their interim analysis process, and the contents of the press release, the case for having serious doubts about those results is strong. We certainly don’t want to throw babies out with the bath water. Just because the trials and data analyses are badly flawed, doesn’t mean that the vaccine is too. It just means that the US trial – including any new dosing regimen they want to test – has to be rigorous, and it’s indispensable. What’s just happened has the hallmarks of being “Plan B” after the big US trial they talked about in July took so long to get off the ground – only to be grounded for weeks because of safety review. But “Plan A” shouldn’t all that far off.

Meanwhile, the mRNA vaccines from BNT/Pfizer and Moderna are well down the road. Results for 3 inactivated virus vaccines are just around the corner: 1 from Sinovac, and 2 from Sinopharm. And it probably won’t be all that long till we start hearing more about Johnson & Johnson’s single shot vaccine.


Hilda Bastian

Originally posted on November 25, 2020

Update November 26: Added link on severe Covid-19.  

Disclosures: My only interest in Covid-19 trials is as a person worried about the virus, as one of my sons is immunocompromised. I have worked for an institute of the NIH in the past, but not any of the institutes working on Covid-19 vaccines. More about me.


The cartoon at the top of this post is my own (CC BY-NC-ND license).


Notes on my collection of clinical trial records on Zotero 

This is a publicly accessible collection I update regularly. It includes any Covid-19 vaccine with published preclinical, clinical trial results, or trial protocols. If a phase 3 trial starts (or is about to) without any prior publications, that trial would also be included. Once a vaccine is in the collection, clinical trial register entries for that vaccine are also added.

When trials are registered in more than clinical trials registry, the multiple records may or may not be in the collection: If I have located a record in ClinicalTrials.gov, I do not hunt for additional registrations. For my own convenience in keeping an overview of vaccine progress, when preprints appear later in journals, I over-write the original record with the journal article.

The entries are tagged by vaccine and what type of record it is – phase 2 results, or a trial protocol, for example. Those tags are bottom left: clicking on any of the record types will narrow down the list of records to those types of records and/or specific vaccines. Double-click on any of the actual records, and it will take you to it. Clicking off a specific vaccine’s tag will get you into the full collection for all vaccines with published results and/or phase 3 trials. For example, this link shows all the protocols I know of.

You can notify me of anything missing either via Twitter or via a comment on one of my Covid-19 vaccine monthly updates at PLOS. (I moderate comments at the blog, so if you want your message to stay private, just say so.)



Find me on: