Identifying if Pfizer had bad batches from a lot size - Freedom Of Information Act (FOIA) request
This time we have the denominator (5 mins)
ICAN (Informed Consent Action Network) has just released data on the number of doses per Pfizer lot. This means we can now calculate a ‘death rate’ on VAERS.
I’ve written before about the deaths in VAERS. The article below, for example, showed that the time between injection and death among those aged under 40 was half that of those over 40.
I’ve also analyzed bad batches before and found my results confirmed the work done by Project Enigma who analyzed lot consistency and potential breaches of GMP (Good Manufacturing Practice). Whilst collaborating on that work, I wanted to calculate a death rate to normalize the results. So I needed to know the lot sizes and at that time (a few months ago) this was not available.
But no more …
FOIA
Aaron Siri and ICAN recently received a response to their FOIA asking for Pfizer lots sizes. And guess what? They got it for the US (as Aaron claimed on The Highwire) and have published it.
After some tidying and cross-checking with Sasha Latypova (who probably knows more about bad batches than anyone), I managed to distill 157 lots of interest.
Deaths by the manufacturer (VAERS)
I downloaded the latest domestic VAERS data from CDC recently (to 07-18-2022), which I considered good enough for this analysis. After filtering the data as follows, I began looking at deaths by all manufacturers
Filters:
None = 13,705 deaths
VAX_LOT_NUMBER <> NULL = 10,153 deaths
VAX_LOT_NUMBER <> ‘UNKNOWN‘ = 9,714 deaths
From the graph below we can see that Pfizer and Moderna are tracking each other with Pfizer having 4,508 deaths and Moderna 4,334. Note that in VAERS some reports of death are not associated with a valid lot number. For this analysis, we are only using entries with valid lot numbers.
Deaths by Manufacturer: Source - VAERS
Pfizer lots
Let’s filter out just the Pfizer lots now. Doing so, we find that the 4,508 deaths attributed after the PFIZER/BIONTECH vaccine code in VAERS are spread across an eye-watering 469 lots! That’s because lots are entered manually in VAERS. Some get entered lowercase, some upper case, some with symbols like ‘#’ at the beginning and some just have typos in them making all of the above unreadable.
So let’s convert them all to upper case and match each lot to the denominator given in the FOIA data that we now have from Pfizer. This reduces our deaths to 4,055 mapped to just 148 lots. So we’ve lost 453 VAERS records (deaths) or ~10% but have matched this 90% of total deaths onto just 148 lot numbers from the FOIA.
We can now calculate a death rate using the numerator (i.e. the number of deaths in VAERS) divided by the number of lots shipped in the FOIA (the denominator).
For example, if lot ABC has 200 deaths reported and (from the FOIA data) there were 20,000 doses for this lot: 200/20,000 = 1%. Thus at least 1% (or more) of the doses in this batch caused death. We say at least 1% or more because possibly not all 20,000 doses were administered (they were recorded as ‘shipped’).
Pfizer deaths by lot number: Source - VAERS and FOIA
This is the first time we’ve been able to plot a death rate with a denominator from the manufacturer. Look at what it tells you. The x-axis is the lot number ordered by alphanumeric sequence and the y-axis is the death rate / 10,000 doses ‘shipped’.
For Pfizer, we show below that lot numbers arranged by alphanumeric strongly correlate with the order in which the lots were administered. This demonstrates that initial death rates in the rollout were much higher than later on.
The total number of doses administered by Pfizer according to Statista today was 355M. The total number of doses shipped in the FOIA was 425M. Let’s assume there are 70M doses in circulation in the US today (vaccinations are falling meaning stock is piling up) so I think the FOIA volumes pass the ‘red face test’. Tick.
But look at the graph again, there’s a clear ‘impulse’ to the left of the plot (ignore the first points in the series that start with a ‘3’ - we’ll address these shortly). The data indicate that the death rate between EJ1685 and EN6204 was around 0.5 per 10,000 doses or 1:20,000 doses. That isn’t a million miles away from what others have calculated here in the UK from the number of deaths reported on the Yellow Card. So another ‘red face test’ passed. Tick again.
But see how this curve tails off almost predictably from EN6204 to the end of the plot (note - I am plotting every 2nd lot number for ease of visibility).
What’s going on?
So what explains this shape? There are a few potential answers which I’ll put forward. You may consider others yourself, but let’s put these up for now:
It was due to older people dying early in the rollout
It was due to manufacturing changes/improvements
It was due to an adulterated product
All three would need to be supported by the assumption that lot number order was related to chronological order. And here’s where VAERS comes unstuck.
I have previously discovered that VAX_DATE in VAERS gets set when a patient first records an adverse event with, for example, dose 1. This is then not updated if they suffer another event in dose 2 (or 3..n). This means the only reliable way to ascertain when a lot number first appears in VAERS is when considering reports from (and only from) those who have received dose 1.
We can use all event data related to Pfizer with dose 1 responses which produce a set of 93,568 records to begin with when combined with the FOIA data.
If we now number the lots in alphanumeric order 1..m and the number the dates in chronological order 1..n and then plot these, if there’s a relationship between them we’ll see it. And we do.
Lot order number plotted against date order: Source - VAERS and FOIA
Immediately, we can see that the lots beginning with “3” do not fit into the beginning of the sequence, so let’s remove them and replot the graph (though you could slide them to the right if you really wanted to).
Lot order number plotted against date order (adjusted): Source - VAERS and FOIA
That’s better and a pretty tasty R^2 of 0.83 along with a correlation coefficient of 0.91, so I’ll assume the lot order number is related to time then. Good.
Lastly, turning our attention to ‘why’ there’s an obvious increased death rate at the beginning of the rollout, I am going to avoid that myself but am interested to hear your thoughts.
What do you think? Vote or leave a comment below.
Hi! Thanks for the analysis. Very informative. Have you seen this article by Steve Kirsch? Perhaps the deaths went down over time because the vials don't contain mRNA? I'm curious to know how many severe adverse events are associated with later lots.
https://stevekirsch.substack.com/p/want-to-know-whats-inside-the-vaccine
Thank you for the research, great article and info!
My following speculations are in response to your question about "why" we think there may be a strong signal in the beginning of the rollout. In thinking about the injectables from a nanomaterial and R&D perspective, it would be pertinent to know the consistency of nanoparticle sizes from lot to lot. Properties of nanoparticles can be profoundly different when comparing a 10nm vs 100nm particle due to active surface area alone - hence, 0.3 mcg of 10nm particles can have substantially different response compared to 0.3 mcg of 100nm particles. Also, nanoparticles can be unstable and the storage conditions of these genetic products seem to suggest that they are in fact unstable when compared to traditional vaccines. Nanoparticles tend to agglomerate when they are unstable leading to large particle sizes which is one of the reasons why the solution can go from clear to hazy - the larger particles resulting from agglomeration are finally large enough to reflect light causing us to see the haze. Agglomeration whether from bad GMP, storage and/or handling could potentially make the injectable less potent simply due to the fact that the agglomerated lipid nanoparticles (essentially drug delivery vehicles) are not protecting and delivering the intended optimum dosage of active mRNA. I often find myself wondering if early roll-out lots were experimental formulations which got tweaked when they got the "oh sh$t" signals.