Thread by @hildabast

EMA (Euro drug regulator) assessment report for the Oxford-AstraZeneca vaccine is online. This one's 181 pages. Most of it's based on the original November data (as in the Lancet publication) but some includes is Ded 7 data (as in recent preprint): ema.europa.eu/en/medicines/h… ...1/n

...They had a major objection to the US manufacturing plant (it's also manufactured in the UK & Belgium - EMA is not assessing Covishield, the Indian version, at all). However, their concerns were resolved. Sounds like it was missing certification...2/n

...EMA wants additional data to validate quality of manufacture at all 3 sites is comparable. Since there's no assessment of Covishield here & WHO didn't do it, I assume this rests totally on the Indian regulator, which published no data (Indian trial still unpublished)...3/n

...Given the issues with manufacturing in the trial & early on, it's reassuring to see this assessment of manufacturing quality. (Not my area of expertise.)...4/n

...They include some data from a preliminary study in mice on reproductive safety: the main study isn't finished yet. "The data do not reveal causes of concern regarding safety". The product is not known to pass on through milk...5/n

... Here's the explanation of what efficacy data is in & not: it falls in between the data in the original Lancet publication & the recent 2 preprints. (COV001 is UK phase 1/2, COV005 is South Africa phase 1/2)...6/n

...The number of people 65 & older increased in the dataset between November (1st pic) & December (2nd) (only standard dosing in the December set) but some data in December hadn't been fully processed. Efficacy analyses were based on Nov dataset ...7/n

..This set of efficacy analyses included data from Dec dataset: footnote a.

Overall standard dose efficacy: 62.6% (CI: 51-72)
Standard dose, 4-12 week interval: 59.5% (CI: 46-70)

Low bounds of CIs for individual trials shown: protocol stipulated min 20%, not 30% like WHO...8/n

...Why EMA excluded low dose group: "As the SD/SD is the intended regimen to be used in real life, the estimated VE based on the SD/SD seronegative at baseline population are expected to provide a better approximation of the expected efficacy". ...9/n

...For that December standard dose efficacy analysis, there was only 1 person with severe Covid-19 (in the control group), 0 people in intensive care & 0 deaths. 0 people in the vaccinated group were hospitalized for Covid-19 vs 8 in the control group...10/n

...That all shows how they got to their final estimate of efficacy of 60% for symptomatic Covid-19, details shown here (based on the December 7 analysis)...11/n

...On the question of asymptomatic infection: "Numbers are small and do not provide sufficient evidence to make conclusions regarding the efficacy of AZD1222 against
asymptomatic SARS-CoV-2 infection"...12/n

...As older people were added later in the trial, they were getting their second dose in less than 6 weeks. Efficacy for 65 & older: 44.8% (CI: -89 to 84). There were too few people aged 46 to <65 to analyze. Covid-19 illnesses in that age group: 8 (vax) vs 9 (control)...13/n

...There were too few people who had previously been infected with SARS-CoV-2 to analyze. The different dosing intervals gets confusing, as it's partly Nov & partly Dec data: confidence intervals still very wide though (=high uncertainty)...14/n

...(Dosing intervals table is on pages 106-107 if you want to look them up, by the way.)...15/n

...They conclude, despite the problems they identified, the data are "sufficiently robust to allow conclusions" - those conclusions follow ...16/n

..60% efficacy; protects against severe illness, hospitalization, but numbers too small to estimate how much; can infer protection in people over 55, but there's not enough data. They want to see the US/LatAm trial for older people & those with underlying health conditions...17/n

..This is new: serious adverse events (SAE) of interest in these trials + the US/LatAm study so far & Indian trial SAE too. They conclude some of the neurological SAEs are unrelated to the vax, but in some, couldn't conclude one way or the other...18/n

...Further on the SAEs: no specific risk identified. They conclude neurological adverse event should be closely monitored as part of pharmacovigilance, & given auto-immune events, wants to see some more background on the viral vector & human neurological tissue... 19/n

...Overall, their conclusion on the quality of the evidence: "The conduct of studies was sub-optimal with regards to substantial changes to the protocol made after the start of studies, errors in dosing & an unplanned varying dose interval between 4 & 26 weeks..../20n

..."Adaptations to confirmatory trials introduced without proper planning reduce the confirmatory nature of the trial." Yes, agree. Further confirmation of just how critical the US/LatAm trial of this vaccine is. (Results expected in a couple of weeks or so.) /21