All the way back in June 2021 I worked with a small group of amazing people that included some medical doctors, a nurse, a midwife, mathematicians, statisticians and computer (data) scientists to produce an interim report on the first 250 post-covid-19 vaccination deaths reported in the Vaccine Adverse Event Reporting System (VAERS). That interim report can be found HERE, and it was downloaded from ResearchGate almost 200,000 times, cited in numerous other academic publications, and discussed favourably by prominent academic clinicians like Dr Peter McCullough on several occassions.
Many wonderful and inquisitive souls contacted me to ask when I would be following that report up…
…and while I did actually do the work during June and July of 2022 for the first annual follow-up results batch, it got ‘lost in committee’ when I sent it out to the potential collaborators in our wider research group who, in fairness, all had complex and important Covid-19 related work that they were already doing (amongst others, this group includes the inimitable Dr Jessica Rose, Jikky the Mouse, Jonathan Engler, Professors Norman Fenton and Martin Neil, Josh Guetzkow, Joel Smalley and Dr Clare Craig). As a result, and I apologise to you all, it got forgotten as even I went on to other things and never actually managed to release it.
I am now going to correct this omission.
Readers of my substack will today get ‘the scoop’ ahead of my announcing release of the preprint report that everyone deserved to have had 6 months ago.
Background: The definition of vaccination, vaccines and vaccine adverse event reporting
Scientific disagreements persist concerning almost every aspect of SARS-CoV-2, Covid-19 and the global response. Of all the contested issues surrounding Covid-19 none has become more contentious than that of the Covid-19 vaccinations. Disagreement exists on whether: (i) they are effective at preventing transmission and blocking infection (Stokel-Walker, 2022) or not (Singanayagam et al, 2022); (ii) they have saved lives or not; and (iii) even whether or not administration of the mRNA-based injections fits within the blanket term vaccination such that the CDC is said by some to have altered the definition in order to make them qualify.
It is important to understand the motivation for why we do something. There is a vast difference between doing something I am told is both low risk and preventative for me catching or dying from a known deadly disease, versus doing the same thing with even an assumed minor risk to me but where the disease profile means my own life was never really in danger from the disease - and the intention has been framed as protecting some other person at some future date. The definition of vaccination focuses on the intended purpose for the vaccine - in effect, why we are being given (or consenting to receive) the injection. When we look at the intention of early vaccines such as Jenner’s cowpox variolation to inoculate against smallpox or Pasteur’s rabies vaccine, the purpose was demonstrably to prevent illness in the recipient by exposing them to an attenuated or denatured version of the infectious organism, thus protecting them when they later become exposed to the wild virus (CDC, 1985; Reidel, 2005). This disease prevention intention sounds more resolute and makes the why of vaccines an easier sell, and vaccination by this approach of mild exposure is observed in even recent vaccine studies (Nigrovic & Thompson, 2007) and the CDC’s pre-2015 definition for the term vaccination shown in Figure 1.
As can be seen, the CDC definition for vaccination has changed the purpose from an intention to prevent, to one which seeks to produce immunity in 2015, and more recently in 2021 at the height of the global Covid-19 vaccine rollout to one intended to produce protection. While the CDC suggest that these slight changes in wording are innocuous and haven’t impacted the overall definition (Camero, 2021), when we consider the terms that have been changed and the way the CDC definition is couched, there appears a strong inverse relationship with the tolerability of failure of the vaccine (i.e. it simply failing to do what it should). An intention to prevent disease is significantly stronger and less tolerant to failure than one that will only produce immunity, which in turn was stronger than the latest that will only produce protection. With each change the strength of purpose is weakened - and this can even be seen in how vaccine failure is now discussed in the media and hence, perceived by the public.
It is also important to understand how issues that arise from doing something will be monitored, and who will be responsible for managing the monitoring system and making decisions regarding the data that must be collected. As with the previous analysis (McLachlan et al, 2021), the work presented here is helped by the fact that the United States of America (USA), in contrast to other countries, provides granular datasets for post-licensure surveillance of vaccines through their publicly accessible VAERS database. In contrast, while transparency is promised, the pharmacovigilance approaches of the United Kingdom (UK Yellow Card System), Australia (AUS Adverse Event Management System), and New Zealand (NZ Adverse Event Reporting Form) still tend towards releasing only summarised statistics. These cover patient age, symptoms and in some cases, comorbidities, but do not enable reviewers to determine whether each symptom or comorbidity occurred alone, or whether there are underlying relationships between age, symptoms and comorbidities that may be more likely to produce a severe adverse event. The UK Government’s own advice on what to include in your yellow card of an adverse event reaction does not stipulate inclusion of vital information such as current illnesses, other medications being taken, and any diagnosed comorbidities. These items are considered later, and only as additional information - in language couched to suggest they are supplementary and thus belying their potential importance. This means that key data that might lead to detection of a safety issue with a new medicine is not systematically collected and even when it is, in contrast to the promises of transparency the reporting formats of the UK, AUS and NZ regulators, this means that it is not being made available to the public in a form that enables analysis and consideration. It has been known since 2004 (Miller, 2004) that national adverse event reporting systems are clearly inadequate, or else fallacious, and so are unfit for the purpose of providing reliable records for analysis of the incidence of vaccine-related adverse health events.
However, there is also the issue of what a potential source of monitoring data can tell us, and when a previously acceptable risk may be found intolerable. VAERS was implemented as a tool for post-market surveillance of vaccines. VAERS collects data about adverse events and harmful side effects that occur after vaccine administration. VAERS accepts reports from clinicians, allied healthcare workers and the general public, and it is this last class of reporter that has been strongly suggested as the weakness that makes Covid-19 vaccine-related VAERS data meaningless - open to unverified reports by the general public and fake reports by ‘anti-vaxxers’ (Jarry, 2021; Motta & Stecula, 2021; Wadman, 2021). However, it is known that the CDC prunes those reports that their staff consider to be fake (Wadman, 2021), and while this should alleviate at least some of the claims that ‘anti-vaxxers’ are polluting the data with false reports, it has also resulted in real reports by healthcare workers also being removed.
Some advocates for vaccine makers have even claimed that vaccine injury claims do not even have to be scientifically valid to unsettle the pharmaceutical giants who we are deceptively led to believe are not making very much money from their vaccine endeavours (Jarry, 2021). For example: (i) Pfizer have reportedly made more than $37bn from sales of the Covid-19 vaccine; (ii) GSK made more than $25mil from LYMErix sales in the year it was on the market (2000), and more than $842mil from the four vaccines they sold during that year; and (iii) Moderna - a company that had previously never offered a product for sale, has already reported an $18.5bil profit on the approximately $1.4bil they received from their government, taxpayer and philanthropist-funded Covid-19 vaccine. These amounts can hardly be described as not very much money. Leaving aside profit, we resolved the issue regarding whether ‘anti-vaxxers’ were making false reports against vaccines and hence tainting VAERS in our first study when, on analysis of the textual narrative for each report, we were able to conclude that at least 72% of the 250 reports we reviewed were authored by clinicians, allied health workers or the staff of pharmaceutical companies that had manufactured the vaccines (McLachlan et al, 2021). While the remaining 28% were adjudicated to have been made by lay persons (family and friends of the deceased), 3% of the 250 total reports we reviewed - all coming from those we judged to have been written by lay persons, showed evidence that the VAERS call centre employee (the recorder) had provided their own narrative, interpretation or commentary on the information that had been provided by the lay person (the reporter). So, if the reporter is not as significant an issue as journalistic fact checkers would have us believe, the question becomes what can we learn from careful and cautious consideration of the VAERS dataset? The results of this second review will update all of these statistics, and will further demonstrate whether the majority of ongoing VAERS reports continue to be made by healthcare professionals or lay persons.
VAERS data analysis has previously shown that vaccines may be over-activating the immune system and several are linked to incidence of a condition called Polymyalgia Rheumatica that is currently the subject of several class action lawsuits in the United States of America (Bassendine & Bridge, 2020; Falsetti et al, 2020; Liozen et al, 2021; Manzo et al, 2021). VAERS data regarding more than 64,000 cases of adverse events, including 547 deaths, has been shown to demonstrate a link between HPV/Gardasil vaccination and the incidence of AEs at different rates in different ethnic groups (Huang et al, 2018), and has been presented as evidence in the complaints of as many as seven lawsuits filed against Merck and Kaiser and their HPV/Gardasil vaccines (including Brunkner v Merck anors in the Superior Court of the State of California, Sullivan v Merck & Co in the Superior Court of New Jersey, and Dalton v Merck & Co in the United States District Court for the Eastern District of Michigan). The complaints in these cases also allege that Merck, rather than ‘anti-vaxxers’, has submitted fraudulent reports to VAERS to conceal Gardasil’s link to the deaths of teenagers. Finally, research has demonstrated that VAERS data can be used effectively to detect systemic vaccine adverse events (Wang et al, 2018).
Other things we know about VAERS include that several government investigations have resolved that vaccine adverse events are considerably underreported in the VAERS dataset, that concerns regarding instances where clinicians were discouraged from reporting post-vaccination health events that resulted in lost days of employment or hospitalisation were founded, and that some administering clinicians did not report to VAERS for fear of potential legal liability for the adverse event following vaccination (IOM, 1997; Jollenbeck et al, 2002). And finally, that even the US Congress’ General Accounting Office found that vaccine manufacturers misleadingly understate the potential for adverse events - with the military Anthrax vaccine deployment demonstrating a significant adverse event rate almost 200% higher (85%) than the manufacturer’s claim (30%) (Roos, 2002).
In the late 1990’s and just like we have seen for the Covid-19 vaccines, the GlaxoSmithKline (GSK) LYMErix vaccine received significant media attention both because it failed to produce any immunity in more than 20% of recipients, and more significantly because post-marketing monitoring data linked it to incidence of long term and chronic disease for a small number of recipients who became known as vaccine victims (Nigrovic & Thompson, 2007; Sheller, 2013). Like many have maintained about the Covid-19 vaccines, LYMErix was rushed to market without adequate testing and, even more significantly, in spite of strong evidence to the contrary the media then, as now, claimed there was no indication of any long-term adverse reactions (Sheller, 2013). Within a year of its release LYMErix was voluntarily withdrawn by GSK. The public simply could not tolerate a vaccine that not only did not prevent disease in a large group of recipients, but one which was also found to produce adverse reactions in small groups of recipients. The CDC’s more recent shift from vaccines for disease prevention to vaccines to produce protection has allowed pharmaceutical company and media narratives to adapt as it has become clear that a previous claim about the Covid-19 vaccines was no longer a fair representation of what was actually happening. This has seen the Covid-19 vaccines go from being pronounced as 90-95% effective at preventing Covid-19 disease (Herper, 2020) to their being described as 80% protective against severe disease or death, and more recently, simply as preventing severe illness (Benmeleh, 2021). As with the CDC definition for vaccination, with each change in headline the purpose for giving (or receiving) Covid-19 vaccination has weakened.
As with the example of LYMErix, determination of their safety is actually the most pressing issue - and is the primary focus of this report. In particular, we consider whether the observed safety signal is significantly stronger than that which has historically triggered regulators to withdraw or recall medicines or vaccines from the market that were deemed unsafe for continued use. There may already be sufficient data reported to demonstrate that the costs (based not just on the cost-per-dose for the billions of doses already administered, but also on the negative impacts to recipient’s health) outweigh the benefits (prevention of an individual becoming infected with the target disease and transmitting it to others) and suggest we should pause and reconsider their continued administration.
In conclusion: We have seen that the definition for vaccination has been slowly weakened such that their intended purpose has gone from the more qualified goal to prevent disease in the recipient, to a less quantifiable goal to produce protection. This has allowed the public perception of vaccination in the mainstream media to become a moving feast - that during covid has seen their headline efficacy goal change from being measured as preventing infections to a less provable preventing severe illness. We have seen that other vaccines have been withdrawn or recalled when they: (a) have failed to produce immunity, and hence failed to prevent disease in a large segment of the population; and (b) had been linked to small clusters of adverse events. We also saw that while fact checkers claimed that VAERS data is plagued by fraudulent reports by anti-vaxxers and unverified reports by the general public, the majority of reports in our previous analysis were made by healthcare workers and hence, were more likely to be credible. Other claims made by fact checkers, including that the pharmaceutical companies were not making significant profits from vaccines, were also seen to be incorrect. Finally, we saw that VAERS data: (i) has previously been used to identify clusters of adverse events and a link between particular vaccines and the incidence of autoimmune inflammatory disease; (ii) is capable of supporting research to identify systemic adverse events caused by vaccines; and (iii) that complaints filed in lawsuits against vaccine manufacturers allege that it has actually been the pharmaceutical companies who may have misused the VAERS system by entering fraudulent reports.
Method: What did I do differently the second time around?
On the 3rd of April, 2022 I downloaded the 2022 Full VAERS Zip File from the HHS VAERS website. Like all VAERS files, this file contains three comma-separated value (.CSV) files each containing a single dataset: (i) VAERS data; (ii) VAERS symptoms; and (iii) VAERS vaccine. The VAERS data contains basic demographics (age, US State), dates (vaccination, symptom onset, death) and a textual (and often clinical) narrative of the report contained within a column called SYMPTOM_TEXT. The VAERS symptoms data lists up to five symptoms experienced by the subject patient and which it is suggested are related to the current presentation that involves or resulted from the adverse event arising out of injection of a Covid-19 vaccination. Finally, the VAERS vaccine data provides specific detail regarding the vaccine administered (batch/lot number, brand name, manufacturer), now including the dose number for that vaccine series (first, second, third), and administration route (intramuscular, intravenous, subcutaneous). An individual VAERS_ID identifier number is assigned to each subject patient reported to VAERS and is used across each of the three data files to identify information relating to that individual. This enables us to reconstruct the three data files into a single contiguous dataset for use in our analysis.
In our first analysis we relied on the manual parsing of the symptom text and symptom lists for each individual VAERS subject by at least two reviewers - of which at least one had to be clinical. Each clinical narrative was broken down into its qualitative elements, quantitatively coded and verified. This was a time consuming process that took several weeks to complete. In this new analysis I began with a larger potential pool of records to review, having undertaken to review 1012 of the reports involving death that presented with complete or near-complete responses in every column. I had to work smarter, not harder. Machine Learning Text Classifiers (MLTC) were trained initially on the dataset developed out of our previous VAERS review, as well as with an initial subset of the first 100 of the 1012 records being reviewed from the 2022 data. This work was performed by the lead author (a computer scientist and health informatician previously trained in undergraduate pre-registration clinical nursing) and one currently practicing clinician. The MLTC were used to perform the first pass on the remaining 912 reports - identifying, classifying, and coding information, and highlighting elements that the classifier found to be inconsistent or deemed inconclusive for human review. The human review process was not only tasked with resolving these highlighted items, but also performed manual validation on a random subset of 200 records. The manual validation process allowed for identification and correction of any potential inaccuracies of the method, fine tuning of the classifiers, and ensured the same degree of overall accuracy as the previous interim report. Some of the areas where this improved both the qualitative and quantitative accuracy and overall speed include where multiple terms and acronyms are used to describe the same clinical element. For example, the clinical notes of individuals with hypertension might mention one or more of the following: (i) hypertension; (ii) high blood pressure; (iii) HBP; (iv) BP^; (v) HTN; (vi) elevated systolic pressure; or (vii) elevated blood pressure. It is also possible to include pre-eclampsia toxaemia (PET) or pregnancy induced hypertension (PIH) in the context of a woman who is pregnant. In this example, and context aside, there may be seven, nine or possibly even more ways to report what is fundamentally the same measured symptom. This overall approach allowed for processing the 1012 records while expending around the same time and effort as had been applied in the previous VAERS data review of only 250 records.
Summary Results: The key differences and new realisations from this second (2022) cohort include that:
As the Covid-19 vaccines were rolled out sequentially to the decreasing age groups, the percentage of deaths reported in those groups were seen to increase.
Disparity in the likelihood of a reported death based on gender has significantly reduced (deaths in males in the original 2021 cohort were 1.6 times more likely based on the reviewed reports).
All reported deaths in minors reviewed in the 2022 cohort (those below the age of 15 years) were males, and in most age groups males were still more frequently reported.
The majority of VAERS death reports continue to demonstrate characteristics strongly supporting that a healthcare provider was the reporting person.
While 31% fewer people in the 2022 cohort had reported comorbidities, the incidence of cardiac, hypertensive and diabetic comorbidities was generally consistent with the 2021 cohort.
For the first time the CDC recognised the occurrence of post-vaccination breakthrough infections which were described in the reports of almost half (49.1%) of the 2022 cohort.
For the complete analysis and to view the comprehensive results you can download the preprint of our VAERS 2.0 Results report HERE.
There are legal and ethical reasons why VAERS data is important and cannot, and should not, be dismissed: (1) it is the declared and legally required pharmacovigilance of the US government. This is supposed to be a part of a comprehensive legal schema passed by Congress that takes away the individual right to sue over vaccine harms and in return created VAERS specifically to detect safety signals in the population. (2) Healthcare providers are required by law (by the FDC&A and the FDA regulations in the CFR) to report AEs to VAERS. (3) Government officials should be legally estopped from arguing that VAERS isn't reliable. You don't get to pass a law that requires a pharmacovigilance system for safety and then claim that system is unreliable when it starts to look really bad. (4) Particularly not where for decades we've used VAERS, and the government has used it to claim how safe other vaccines are because of their low reporting for LITERALLY EVERY OTHER VACCINE in the country. This is as dishonest and unethical as it's possible to be.
I would be interested in any stored recoding or look-up tables that you may have created to group similarly named symptoms. Are those available?