Digging Deeper Into EudraVigilance (part 4)
A quick recap on suspected drugs
The datafile field headed Suspect/interacting Drug List (Drug Char - Indication PT - Action taken - [Duration - Dose - Route]) is where all suspected and interacting drugs are listed, many of which include COVID vaccination products. We have established that out of 23,891 fatalities over the period 20 Dec 2020 - 28 Feb 2022 some 21,799 (91.2%) possessed records with at least one suspected COVID vaccination product listed, with Pfizer notching-up 2,085 zero suspected reactions and AstraZeneca notching-up 7 zero suspected reactions.
Quite how some shots come to be ignored when the vast majority are not is most curious. One thing I have noted is that those 2,085 non-suspicious Pfizer shots occur in a regular manner throughout the sample period, so its not a case of data capture flows settling down but an ongoing issue. I thus decided to don my deerstalker and have a rummage…
The Pfizer Enigma
It transpires there are no less than 7 interacting factors that influence coding of Pfizer’s product as a suspected drug and we shall walk through these in quick succession. A total of 13,354/23,891 (55.9%) fatalities were calculated to have been in receipt of one or more Cominarty doses and this sample forms the basis for the following analyses, the idea being to identify patterns in distribution for those 2,085 unlisted shots.
Economic Region
There is a much greater reluctance to list Comirnaty as a suspect drug within the non-European Economic Area, with a crude odds ratio of 7.4 (p<0.001; Fisher’s Exact Test, n=13,354), this result revealing the potential for substantial regional reporting bias in a multi-nation database. Please refer to EudraVigilance for regional definitions.
Primary Source Qualification
There is a greater reluctance for healthcare professionals to list Comirnaty as a suspect drug compared to non-healthcare professionals, with a crude odds ratio of 1.7 (p<0.001; Fisher’s Exact Test, n=13,354), this result revealing another source of bias.
Age Group
There is an astonishing and singular reluctance for Comirnaty to be listed as a suspect drug for the 18 - 65y age group. Just what on Earth is going on here?
COVID-19 Diagnosis
There is a reluctance for Cominarty to be listed as a suspect drug for those not suffering from COVID-19 (or at least returning a positive test result), with a crude odds ratio of 2.1 (p<0.001; Fisher’s Exact Test, n=13,354). This result is not entirely unexpected and may offer modest evidence of negative benefits.
Concomitant Drug Use
There is a reluctance for Cominarty to be listed as a suspect drug for those also in receipt of a concomitant drug, with a crude odds ratio of 1.6 (p<0.001; Fisher’s Exact Test, n=13,354). Concomitant drugs are generally those on repeat prescription with historical use for chronic conditions, examples being: Amoxicillin, Salbutamol, Aspirin, Citalopram and Insulin. Whilst this result is in line with expectation it does raise concern over yet more reporting bias.
Total Drug Entries
There is a strong association between the total drug entries made on a patient record and reluctance for Cominarty to be listed as a suspect drug (p<0.001; Mann Whitney U-Test, n=13,354) such that healthcare and non-healthcare professionals would appear reluctant to add yet another suspect drug to a list of suspect drugs. We appear to have unearthed another source of reporting bias!
Total Reactions Recorded
There is a strong association between the total reaction entries made on a patient record and reluctance for Cominarty to be listed as a suspect drug (p<0.001; Mann Whitney U-Test, n=13,354) such that healthcare and non-healthcare professionals would appear reluctant to append Comirnaty as suspect drug to the records of those with more complex conditions. We appear to have unearthed yet another source of reporting bias!
Bias With a Capital ‘B’
This is supposed to be a state-of-the-art EMA clinical database for reporting some pretty serious sh*t, yet I’m sitting rubbing a furrowed brow and wondering what hasn’t been captured, what else is hidden from view and what value we can attach to results. There’s a lot to untangle and we’re fast heading toward multivariate statistical modelling to iron out a few things.
Kettle on!
You write about EudraVigilance: " ...I’m sitting rubbing a furrowed brow and wondering what hasn’t been captured, what else is hidden from view, and what value we can attach to results".
That confusion, missing data, etc were all injected into the system by design. Then it's much easy for them to blame poor data collection and call it garbage.
In VAERS they "missed" very frequently the event outcome (like death, but in the comment field the fatality is very clear), date of death, age, vaccine type, date of vaccination. They also have many months' backlogs of the reports. Also in VAERS, they allow only one entry per person with no further editing. So if somebody ends up in a hospital, ER, etc. creates a VAERS report and then dies days, weeks, or months later the first report stays unchanged and hides the real outcome.
Hence these two biggest adverse effect collection systems are pretty consistent in their attempt to hide the truth.