HomeMy WebLinkAbout07.08.26 Board Correspondence - FW_ United States Senate - Senator Ron Johnson’s Report - Unmasked_ How Biden’s Health Officials Purposely Turned A Blind EyeFrom:Clerk of the Board
To:Mutony, Heather
Cc:Lee, Lewis
Subject:Board Correspondence - FW: United States Senate - Senator Ron Johnson’s Report - Unmasked: How Biden’s
Health Officials Purposely Turned A Blind Eye Toward Covid-19 Vaccine Safety Signals
Date:Thursday, July 9, 2026 4:53:36 PM
Attachments:Senate-PSI-Majority-Staff-Interim-Report-April-29-2026-FINAL.pdf
image001.png
Please see Board Correspondence -
Lewis LeeAdministrative Technician - ConfidentialButte County Administration25 County Center Drive, Suite 200 • Oroville, CA 95965T: 530.552.3326www.buttecounty.ca.gov | lelee@buttecounty.ca.gov
From: lance dreiss <lancedreiss@att.net>
Sent: Wednesday, July 8, 2026 11:11 AM
To: Shared Mailbox Clerk of the Board <pcbs@countyofplumas.com>;
Assemblymember.Gallagher@assembly.ca.gov; Senator.Dahle@senate.ca.gov;
davidhollister@countyofplumas.com; sheriff@countyofplumas.com; District Attorney
<District_Attorney@buttecounty.ca.gov>; Nicolereinert@countyofplumas.com; Kitts, Melissa
<mkitts@buttecounty.ca.gov>; Soderstrom, Monica <msoderstrom@buttecounty.ca.gov>; Durfee,
Peter <PDurfee@buttecounty.ca.gov>; Ronald Owens <ronald@muzzledtruth.com>; Kimmelshue,
Tod <TKimmelshue@buttecounty.ca.gov>; Pickett, Andy <APickett@buttecounty.ca.gov>; Connelly,
Bill <BConnelly@buttecounty.ca.gov>; Teeter, Doug <DTeeter@buttecounty.ca.gov>; Beaudoin,
Jarett <JBeaudoin@buttecounty.ca.gov>; Julie Threet <julie4butte5@gmail.com>; Waugh, Melanie
<mwaugh@buttecounty.ca.gov>; Ritter, Tami <TRitter@buttecounty.ca.gov>; Teri DuBose
<Teri.DuBose@mail.house.gov>; Clerk of the Board <clerkoftheboard@buttecounty.ca.gov>;
Stephens, Brad J. <BStephens@buttecounty.ca.gov>
Subject: United States Senate - Senator Ron Johnson’s Report - Unmasked: How Biden’s Health
Officials Purposely Turned A Blind Eye Toward Covid-19 Vaccine Safety Signals
.ATTENTION: This message originated from outside Butte County. Please exercise judgment
before opening attachments, clicking on links, or replying..
Public Record
https://urldefense.com/v3/__https://www.ronjohnson.senate.gov/wp-
content/uploads/2026/05/Senate-PSI-Majority-Staff-Interim-Report-April-29-2026-
FINAL.pdf__;!!KNMwiTCp4spf!BKAOmqvJGXQPfN4WZPJs__HVrrVl1UYooicsFy6RNqstFcZghdc
fipJCx2NB44SPd9Ir7PrkVMD_8g-36mG9PIVN_YE2zA$
diana dreiss
1
Table of Contents
I. Executive Summary .......................................................................................... 2
A. The FDA’s “Gold Standard” of Data Mining (and its flaws) .......................... 5
i. Background ................................................................................................................. 5
ii. Federal Health Officials’ Reliance on EB Data Mining .................................. 6
iii. EB Data Mining Limitations .................................................................................. 7
iv. Dr. Szarfman’s and Dr. DuMouchel’s New and “Superior” EB Data
Mining Methodology ................................................................................................ 8
B. Locking Down Access to FDA’s Data Mining Analyses ...............................10
C. FDA Officials Acknowledge EB Data Mining Limitations and Continue to
Fail to Make Necessary Adjustments ............................................................12
D. Top FDA CBER Officials Eventually Acknowledge the Effectiveness of the
New Data Mining Model Compared to the Current Methodology ...............13
II. Timeline – Biden Health Officials Ignore Safety System that Can Better
Detect Signals for COVID-19 Vaccine Adverse Events ............................15
III. Glossary ........................................................................................................32
IV. Appendix – Chairman Johnson’s Oversight of COVID-19 Vaccines .....35
2
I. Executive Summary
Documents produced by the Department of Health and Human Services (“HHS”) to the
Permanent Subcommittee on Investigations (“PSI” or “the Subcommittee”) show that beginning
in early 2021, Dr. Ana Szarfman, who at the time was a senior medical officer and safety data
mining developer at the Food and Drug Administration (“FDA”), used an updated data analysis
technique that identified dozens of statistically significant safety signals for adverse events
associated with the COVID-19 vaccines.1 She immediately shared her findings with other FDA
officials, including officials responsible for COVID-19 vaccine safety surveillance, but they
largely ignored her and eventually told her to stop her data analyses.
Specifically, on March 26, 2021, Dr. Szarfman, who worked in the FDA’s Center for
Drug Evaluation and Research (“CDER”), shared a data mining analysis of COVID-19 vaccine
adverse events using a newer methodology that reduced masking.2 Masking is a known data
limitation with FDA’s current data mining method that can inhibit the detection of safety signals
for adverse events.3 Dr. Szarfman’s analysis—which was done in collaboration with Dr. William
DuMouchel, the then-Chief Statistician at Oracle and inventor of the data mining algorithm that
supported FDA’s current data mining system—found “49 examples of extreme masking” with
over twenty of those examples of adverse events now showing a statistically significant safety
signal when adjusted for masking.4 In other words, by using a methodology that accounted for
the masking limitation, Dr. Szarfman and Dr. DuMouchel uncovered approximately 25
statistically significant safety signals for adverse events associated with the COVID-19 vaccines
that were not previously detected by FDA’s current methodology, including sudden cardiac
death, Bell’s palsy, and pulmonary infarction.5
Dr. Szarfman shared similar findings with other senior officials at the FDA in this
instance and at least three other times in April 2021, May 2021, and July 2021 when her and Dr.
DuMouchel’s analyses yielded even more statistically significant safety signals with higher
1 HHS records also credit Dr. Szarfman as someone who “worked to develop the data mining system” that FDA was
using at the time. See PSI-HHS-000008257238; David Wiseman, Signal loss by truancy, masking, and filtering, and
underestimation of potential risks and suspected adverse reactions in the Disproportionality Signal Analyses of
VAERS data associated with COVID-19 pro-vaccines, ResearchGate, Sept. 2025,
https://www.researchgate.net/publication/395382959_Signal_loss_by_truancy_masking_and_filtering_and_underest
imation_of_potential_risks_and_suspected_adverse_reactions_in_the_Disproportionality_Signal_Analyses_of_VAE
RS_data_associated_with_COVID-19_pro at 8-9. Dr. Szarfman retired from the FDA in 2025. The Subcommittee
applied Bates stamps to the records cited in the report that were produced by HHS.
2 PSI-HHS-000008257443-44 (with attachment).
3 Id.; See e.g., David Martin et al., Data Mining for Prospective Early Detection of Safety Signals in the Vaccine
Adverse Event Reporting System (VAERS): A Case Study of Febrile Seizures after a 2010–2011 Seasonal Influenza
Virus Vaccine, Drug Safety (2013), https://link.springer.com/article/10.1007/s40264-013-0051-9; Rave Harpaz et al.,
Signaling COVID-19 Vaccine Adverse Events, Drug Safety (2022),
https://link.springer.com/article/10.1007/s40264-022-01186-z.
4 PSI-HHS-000002293255; PSI-HHS-000008263190-91 (emphasis added). Records show that FDA’s threshold for
determining a statistically significant safety signal was when the lower bound of the reporting estimate (EB05)
exceeded 2.0. See PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-A02D-243AA1844D6E at 166; FOIA
production: https://www.fda.gov/media/184988/download?attachment at 23.
5 PSI-HHS-000008257443-44 (with attachment).
3
values for adverse events, including acute myocardial infarction associated with the Moderna
and Pfizer-BioNTech (“Pfizer”) COVID-19 vaccines, non-site specific embolism and thrombosis
associated with the Janssen (“Johnson & Johnson”) and Pfizer vaccines, dementia associated
with the Pfizer vaccine, and “Death and sudden death” associated with the Moderna and
Pfizer vaccines.6
However, rather than warn the public or hold distribution of the vaccines for further
investigation, records appear to indicate that Biden health officials ignored the statistically
significant safety signals uncovered through the new methodology and were more concerned
about Dr. Szarfman’s efforts rather than her troubling findings. In April 2021, after Dr. Szarfman
emailed a data mining analysis apparently underscoring the limitations of FDA’s current system,
one senior FDA official wrote to his colleagues, “[b]efore we potentially reach out to Ana, we
should meet internally – many considerations not suited to email…”7
By May 2021, senior officials in the FDA’s Center for Biologics Evaluation and Research
(“CBER” – the unit responsible for COVID-19 vaccine safety surveillance), including Drs.
David Menschik, Narayan Nair, and Craig Zinderman drafted an email to Dr. Szarfman, that Dr.
Zinderman eventually transmitted to her, directing Dr. Szarfman to “hold off on creating and
sending data mining reports and analyses.”8 Dr. Peter Marks, a staunch supporter of the COVID-
19 vaccines who at the time led CBER, eventually warned Dr. Patrizia Cavazzoni, the then-
Director of CDER, that Dr. Szarfman’s data mining could “create erroneous conflicts that feed in
to anti-vaccination rhetoric.”9
Although senior officials at FDA instructed Dr. Szarfman to “hold off” on generating and
distributing her and Dr. DuMouchel’s analyses, she continued.10 Dr. Szarfman tried to convince
her colleagues that the data mining method she was sharing was more effective than the method
FDA was using. She wrote to FDA officials in early May 2021 that her and Dr. DuMouchel’s
method “strongly reduces confounding” and that the safety signals in this method are much
higher than the signals detected in FDA’s method, particularly for acute myocardial infarction.11
As highlighted in Chairman Johnson’s report on myocarditis and the COVID-19
vaccines, on May 24, 2021, draft notes from a meeting involving federal health officials
contained the question, “Is [the Vaccine Adverse Event Reporting System (“VAERS”)] signaling
for myopericarditis?”12 The answer stated: “For the age groups 16-17 years and 18-24 years,
6 PSI-HHS-000008258306 (with attachment); PSI-HHS-00008258202-03; PSI-HHS-000002208944-45; PSI-HHS-
000004592364-65 (with attachment) (emphasis added).
7 PSI-HHS-000008251979 (emphasis added, ellipsis in original).
8 PSICOVID_00017246; PSI-HHS-000008251530; PSI-HHS-000008251912-13; PSI-HHS-000001195617-19;
PSI_HHS-000008253450-51; PSI-HHS-000001175745-47; PSI-HHS-000001148712-14; PSICOVID_00017246-47.
9 PSI-HHS-000002213753; PSICOVID_00017246-47. See also, Testimony of Dr. Peter Marks before the Select
Subcomm. on the Coronavirus Pandemic Comm. on Oversight and Accountability, U.S. House of Representatives,
Feb. 15, 2024, https://oversight.house.gov/wp-content/uploads/2024/02/FDA-SSCP-Vaccine-Safety-and-
Surveillance-FDA-Written-Testimony-FINAL-Clean.pdf.
10 PSICOVID_00017246-47.
11 PSICOVID_00017245; PSI-HHS-000002208944-45. In a May 6, 2021 email to her FDA colleagues, Dr.
Szarfman wrote, “Note that the ER05 signals of RPGS are higher than the EB05 signals of MGPS.” Id.
12 Chairman Ron Johnson, Failure to Warn: How Federal Health Agencies Downplayed the Risk of Myocarditis and
Other Adverse Events Following COVID-19 Vaccination, Permanent Subcomm. on Investigations, May 21, 2025,
4
yes.”13 Despite this unambiguous acknowledgement of the myopericarditis signal, Biden health
officials continued to fail to immediately warn the public about the risks of cardiac events
associated with the COVID-19 vaccines.14
In early June 2021, Dr. Szarfman emailed an FDA colleague about ongoing discussions
between the FDA and the Centers for Disease Control and Prevention (“CDC”) about COVID-19
vaccines and “myocardial events and the lack of signals in VAERS and other data resources.”15
She enclosed Dr. DuMouchel’s data mining analysis from a month earlier showing higher
statistically significant safety signals for acute myocardial infarction and noted that she and Dr.
DuMouchel “also detected clear signals for other similar events.”16 Dr. Szarfman wrote, “I am
not astonished that [FDA’s data mining system] was unable to detect these signals.”17
In July 2021, Dr. Szarfman warned FDA officials that her and Dr. DuMouchel’s data
mining method detected “an increased mortality signal with the COVID-19 vaccines,” reminding
her colleagues that she and Dr. DuMouchel use “a method that automatically unmask[s] signals
that remain hidden by other data mining methodologies, including by MGPS [FDA’s current data
mining method abbreviated for Multi-item Gamma Poisson Shrinker].”18 Again, in September
2021, Dr. Szarfman urged senior FDA officials to utilize her and Dr. DuMouchel’s data mining
method underscoring that this method is “much better at unmasking signals than MGPS. It
automatically identifies and corrects for confounders. This is an important function to have,
given the pandemic situation.”19
Records indicate that FDA did not make any adjustments to its data mining methodology
as recommended by Dr. Szarfman. However, Dr. Szarfman continued to advocate for the newer
methodology, eventually co-authoring a June 2022 paper with Dr. DuMouchel and others
published in the journal Drug Safety, which discussed their data mining technique and stated that
masking “is roughly eight times more likely to occur with COVID-19 vaccines than with other
vaccines.”20 In July 2022, Dr. Szarfman emailed this article to then-FDA Commissioner Dr.
Robert Califf noting that her paper describes advances in data mining methodology “especially in
the capacity to unmask hidden signals[.]”21 Upon receipt of this and another article, Dr. Califf
https://www.hsgac.senate.gov/wp-content/uploads/2025.05.21-PSI-Majority-Staff-Interim-Report-Failure-to-
Warn.pdf at 20.
13 Id. It is unclear what data mining system was used to determine that VAERS was signaling for myopericarditis.
14 Id. Chairman Johnson’s May 21, 2025 report discussed how Biden health officials failed to issue a formal warning
about myocarditis, despite increasing cases in VAERS, through the Health Alert Network which would have notified
health care providers about the risk of myocarditis associated with the mRNA COVID-19 vaccines. The FDA did
not announce changes to the labels “to include a warning about myocarditis and pericarditis” for the Moderna and
Pfizer COVID-19 vaccines until June 25, 2021, months after first becoming aware of reports of increased cases of
the adverse events. Id. at 3, 6, 27-28.
15 PSI-HHS-000002208944.
16 PSI-HHS-000002208944-45.
17 PSI-HHS-000002208944.
18 PSI-HHS-000004592364.
19 PSI-HHS-000008254471.
20 Rave Harpaz et al., Signaling COVID-19 Vaccine Adverse Events, Drug Safety (2022),
https://link.springer.com/article/10.1007/s40264-022-01186-z.
21 PSI-HHS-000004461747-48.
5
simply replied, “Thanks. These are good.”22 Documents reviewed to date do not indicate
whether Dr. Califf took any action based on his review of Dr. Szarfman’s article.
Other FDA officials were not so thrilled with Dr. Szarfman’s publication. In August
2022, Dr. Richard Forshee, the then-Deputy Director of CBER’s Office of Biostatistics and
Pharmacovigilance, wrote to Dr. Marks, “we have just learned that Dr. Szarfman is a co-author
on a recently published paper based on COVID-19 vaccines and VAERS. We believe there are a
number of issues with the paper and its findings, and we are discussing how to best respond.”23
It remains unclear based on the records received to date whether FDA officials confronted Dr.
Szarfman about her paper.
A. The FDA’s “Gold Standard” of Data Mining (and its flaws)
i. Background
According to HHS, empirical Bayesian (“EB”) data mining is a “statistical method for
identifying disproportionality (excess of reported [adverse events] for [a] product relative to
other products) in large database[s].”24 HHS records credit Dr. DuMouchel as the individual
who “invented the empirical Bayesian data mining algorithm known as Gamma-Poisson
Shrinker (GPS) and its successor MGPS [Multi-item Gamma Poisson Shrinker], which have
been applied to the detection of safety signals in databases of spontaneous adverse drug event
reports.”25
In January 2021, one month after the FDA issued emergency use authorizations (“EUA”)
for the Pfizer and Moderna COVID-19 vaccines, the FDA and CDC published a Standard
Operating Procedure (“SOP”) document describing how the agencies will perform VAERS
surveillance analyses to identify safety concerns for the COVID-19 vaccines.26 One of the
analyses the FDA and CDC utilized was EB data mining, using the Multi-item Gamma Poisson
Shrinker (“MGPS”) algorithm invented by Dr. DuMouchel. FDA oversaw the data mining
which was conducted through software developed by Dr. DuMouchel at Oracle called Empirica
22 Id.
23 PSI-HHS-000004594929.
24 Vaccine Adverse Event Reporting System (VAERS) Standard Operating Procedures for COVID-19, Centers for
Disease Control and Prevention, Jan. 29, 2021,
https://web.archive.org/web/20210319091240/https://www.cdc.gov/vaccinesafety/pdf/VAERS-v2-SOP.pdf at 16-17.
See also PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-A02D-243AA1844D6E at 160 ; FOIA
production: https://www.fda.gov/media/184988/download?attachment at 22.
25 PSI-HHS-000002293255. See also, William DuMouchel, Bayesian Data Mining in Large Frequency Tables, with
an Application to the FDA Spontaneous Reporting System, The American Statistician (1999), available at
https://www.jstor.org/stable/2686093?seq=1.
26 The FDA issued the EUA for the Pfizer-BioNTech COVID-19 vaccine on December 11, 2020. Moderna received
the EUA for its vaccine on December 18, 2020. COVID-19 Timeline, Centers for Disease Control and Prevention,
https://www.cdc.gov/museum/timeline/covid19.html; Vaccine Adverse Event Reporting System (VAERS) Standard
Operating Procedures for COVID-19, Centers for Disease Control and Prevention, Jan. 29, 2021,
https://web.archive.org/web/20210319091240/https://www.cdc.gov/vaccinesafety/pdf/VAERS-v2-SOP.pdf.
6
Signal (“Empirica”).27 Dr. Szarfman is also credited for developing and helping FDA adopt the
data mining system.28 Records show that FDA’s threshold for determining a statistically
significant safety signal was when the lower bound of the reporting estimate (EB05) exceeded
2.0.29
ii. Federal Health Officials’ Reliance on EB Data Mining
Federal health officials publicly touted the strength of EB data mining as “a more robust
technique” for analyzing VAERS reports.30 Internal emails between HHS officials also showed a
preference for EB data mining over other types of data analyses. In June 2022, Dr. Tom
27 Letter from Jeff Reezek, Centers for Disease Control and Prevention to Sen. Ron Johnson, Permanent Subcomm.
on Investigations, Mar. 14, 2023 (on file with Subcomm.); PSI’s March 25, 2026 document release on EB data
mining: https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-A02D-243AA1844D6E at 161;
FOIA production: https://www.fda.gov/media/184988/download?attachment at 23; David Wiseman, Signal loss by
truancy, masking, and filtering, and underestimation of potential risks and suspected adverse reactions in the
Disproportionality Signal Analyses of VAERS data associated with COVID-19 pro-vaccines, ResearchGate, Sept.
2025,
https://www.researchgate.net/publication/395382959_Signal_loss_by_truancy_masking_and_filtering_and_underest
imation_of_potential_risks_and_suspected_adverse_reactions_in_the_Disproportionality_Signal_Analyses_of_VAE
RS_data_associated_with_COVID-19_pro at 8-9.
28 Robert O’Neill and Ana Szarfman, Some US Food and Drug Administration perspectives on data mining for
pediatric safety Assessment, Current Therapeutic Research (2001),
https://www.sciencedirect.com/science/article/abs/pii/S0011393X01800710; Ana Szarfman et al., Use of screening
algorithms and computer systems to efficiently signal higher-than-expected combinations of drugs and events in the
US FDA's spontaneous reports database, Drug Safety (2002), https://link.springer.com/article/10.2165/00002018-
200225060-00001; David Wiseman, Signal loss by truancy, masking, and filtering, and underestimation of potential
risks and suspected adverse reactions in the Disproportionality Signal Analyses of VAERS data associated with
COVID-19 pro-vaccines, ResearchGate, Sept. 2025,
https://www.researchgate.net/publication/395382959_Signal_loss_by_truancy_masking_and_filtering_and_underest
imation_of_potential_risks_and_suspected_adverse_reactions_in_the_Disproportionality_Signal_Analyses_of_VAE
RS_data_associated_with_COVID-19_pro at 8-9; PSI-HHS-000008257238. One FDA official referred to Dr.
Szarfman as someone who “worked to develop the data mining system[.]” Id. In 2002, Dr. Szarfman was
reportedly awarded “the FDA and CDER Outstanding Scientific Achievement Awards for contributions to safety
data mining.” See Professional Activities, Ana Szarfman, ORCID, https://orcid.org/0000-0001-6680-1423.
29 PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-A02D-243AA1844D6E at 166; FOIA
production: https://www.fda.gov/media/184988/download?attachment at 23. Although FDA officials used 2.0 as
their threshold for a statistically significant signal, an FDA official’s April 5, 2021 PowerPoint presentation before
the Advisory Committee on Immunization Practices appeared to recognize that, “Technically, any [Empirical Bayes
Geometric Mean] value above one indicates disproportional reporting.” See PSI’s March 25, 2026 document release
on EB data mining: https://www.hsgac.senate.gov/wp-content/uploads/2025.05.21-Supporting-Documents-Failure-
to-Warn-Part-08.pdf at 47. As medical researcher Dr. David Wiseman wrote in his September 2025 Preprint article,
because health officials used the higher threshold of 2.0, as opposed to 1.0 which would “technically” indicate a
signal, “[s]ignals were filtered out by an inappropriately high detection threshold.” David Wiseman, Signal loss by
truancy, masking, and filtering, and underestimation of potential risks and suspected
adverse reactions in the Disproportionality Signal Analyses of VAERS data associated with COVID-19
pro-vaccines, ResearchGate, Sept. 2025,
https://www.researchgate.net/publication/395382959_Signal_loss_by_truancy_masking_and_filtering_and_underest
imation_of_potential_risks_and_suspected_adverse_reactions_in_the_Disproportionality_Signal_Analyses_of_VAE
RS_data_associated_with_COVID-19_pro at 3.
30 Letter from Rochelle Walensky, Dir., Centers for Disease Control and Prevention to Sen. Ron Johnson, Sept. 2,
2022 (on file with Subcomm.).
7
Shimabukuro, one of the most senior CDC officials working on the COVID-19 vaccine safety
team, lauded FDA’s use of EB data mining, calling it the “‘gold standard’ for disproportionality
analysis” and stating that it “supersedes” other data mining methods “from the perspective of
generating informative data.”31
iii. EB Data Mining Limitations
Even though EB data mining was considered the “gold standard” by some at HHS for
assessing safety signals associated with the COVID-19 vaccines, its effectiveness was limited by
the statistical phenomenon called “masking.” According to Dr. Szarfman’s 2022 article
published in the journal Drug Safety, masking can occur when “signals for a vaccine of interest
are hidden by the presence of other reported vaccines. This masking effect may in turn limit or
delay our understanding of the risks associated with new and established vaccines.”32
PSI Chairman Ron Johnson described masking in his March 23, 2026 letter to HHS,
writing:33
Essentially, if the baseline data used by federal health agencies for determining the
statistical significance of a safety signal combines the signals of other COVID-19
vaccines, that baseline data will drown out or mask the signals of any single
COVID-19 vaccine. To help explain the effects of masking in layman’s terms,
consider this hypothetical: If 100,000 deaths were reported for Moderna’s vaccine,
100,000 deaths were reported for Pfizer’s vaccine, and 10,000 deaths were reported
for all other vaccines, Moderna’s 100,000 deaths may not look significant compared
to an inflated baseline of 110,000 deaths that is made up of 100,000 Pfizer deaths
combined with the 10,000 deaths from all other vaccines.
In this hypothetical, EB data mining would not detect a statistically significant
safety signal because the true rate of deaths was masked for one product when
compared to a baseline that is inflated by the inclusion of deaths from one or more
products with similarly high death rates.
Another simplified way to hypothetically explain the effects of masking would be
to test the adverse events of hemlock by comparing them to the adverse events of
arsenic. Both may be equally dangerous when each is compared separately against
a non-toxic substance such as saline. But, if the effects of hemlock were compared
against a baseline that combines the data for the effects of arsenic and saline, the
adverse events of hemlock may not result in a significant signal because they have
been drowned out or masked by the baseline data which includes the toxic effects
of arsenic.
31 PSI-HHS-000005235281.
32 Rave Harpaz et al., Signaling COVID-19 Vaccine Adverse Events, Drug Safety (2022),
https://link.springer.com/article/10.1007/s40264-022-01186-z.
33 Letter from Ron Johnson, Chairman, Permanent Subcomm. on Investigations to Robert F. Kennedy, Jr., Sec.,
Dep’t of Health and Human Services, Mar. 23, 2026, https://www.ronjohnson.senate.gov/services/files/CA500350-
195E-472C-9F26-BE93B290B9D9.
8
According to a September 2025 preprint article, medical researcher David Wiseman
identified several limitations with EB data mining, including masking, that resulted
in an underestimation of COVID-19 vaccine adverse events.34 Wiseman pointed
out “FDA’s analysis neglected to correct for masking, where signals for one vaccine
are concealed by signals from other vaccines.”35
As explained further in Chairman Johnson’s March 23, 2026 letter, the Subcommittee is
aware that when reports for certain adverse events are so prevalent, a safety signal may
overcome FDA’s data mining statistical limitations.36 For example, statistically significant safety
signals appeared for ischemic stroke in individuals 65 years and older following injection of the
Pfizer-BioNTech bivalent booster in February and March 2023.37 Likewise, as discussed above,
a signal for myopericarditis was detected in May 2021 for age groups 16-17 and 18-24 years (it
is unclear what data mining system detected this signal).38 Yet, in both of these cases, even after
detecting the safety signal, Biden health officials still failed to immediately issue a formal public
warning about the potential risks.
iv. Dr. Szarfman’s and Dr. DuMouchel’s New and “Superior” EB Data Mining
Methodology
For decades, FDA relied on Dr. Szarfman’s and Dr. DuMouchel’s research into data
mining. In 2004, Dr. Szarfman, who at the time worked at FDA in the Office of
Pharmacoepidemiology and Statistical Sciences, published a paper advocating for the FDA’s use
of the Bayesian data mining system that Dr. DuMouchel invented called MGPS.39 Her paper
noted that FDA was evaluating MGPS as a method to “enhance the FDA’s ability to monitor the
safety of drugs, biologics, and vaccines after they have been approved for use.”40 Records
indicate that by January 2021, the MGPS algorithm served as the basis for FDA’s EB data
mining system.41
34 David Wiseman, Signal loss by truancy, masking, and filtering, and underestimation of potential risks and
suspected adverse reactions in the Disproportionality Signal Analyses of VAERS data associated with COVID-19
pro-vaccines, ResearchGate, Sept. 2025,
https://www.researchgate.net/publication/395382959_Signal_loss_by_truancy_masking_and_filtering_and_underest
imation_of_potential_risks_and_suspected_adverse_reactions_in_the_Disproportionality_Signal_Analyses_of_VAE
RS_data_associated_with_COVID-19_pro.
35 Id. at 3.
36 Letter from Ron Johnson, Chairman, Permanent Subcomm. on Investigations to Robert F. Kennedy, Jr., Sec.,
Dep’t of Health and Human Services, Mar. 23, 2026, https://www.ronjohnson.senate.gov/services/files/CA500350-
195E-472C-9F26-BE93B290B9D9 at 15.
37 Id. at 16.
38 Chairman Ron Johnson, Failure to Warn: How Federal Health Agencies Downplayed the Risk of Myocarditis and
Other Adverse Events Following COVID-19 Vaccination, Permanent Subcomm. on Investigations, May 21, 2025,
https://www.hsgac.senate.gov/wp-content/uploads/2025.05.21-PSI-Majority-Staff-Interim-Report-Failure-to-
Warn.pdf at 20.
39 Ana Szarfman et al., Pharmacovigilance in the 21st century: new systematic tools for an old problem,
Pharmacotherapy (2004), https://accpjournals.onlinelibrary.wiley.com/doi/epdf/10.1592/phco.24.13.1099.38090.
40 Id. at 1.
41 See, e.g., PSI-HHS-000008254390-92; FOIA production:
https://www.fda.gov/media/184988/download?attachment at 2-5.
9
On March 1, 2021, just over one month after the publication of the SOP that named FDA
as the agency in charge of using EB data mining to identify statistically significant safety signals
associated with the COVID-19 vaccines, Dr. Szarfman briefed CBER Director Dr. Marks and
other senior FDA officials about a new, “state of the art” EB data mining algorithm, also
invented by Dr. DuMouchel, called Regression-Adjusted Gamma Poisson Shrinker (“RGPS”).42
In her presentation, Dr. Szarfman described RGPS’s performance as “superior” compared to
MGPS, noting that it “can better adjust for both, masking (false negatives) and confounding
(false positives).”43 She emphasized that the “MGPS data mining method currently in use at the
Agency and at the CDC is not state of the art” and that “RGPS is the state of the art” because it
“incorporates more information into the signal generation process. This leads to a lower rate of
missed signals and less false alerts.”44 Despite Dr. Szarfman’s advocacy for a new and
apparently more reliable data mining method, FDA failed to take her expert advice and adopt the
“state of the art” method.45
It is somewhat ironic that Drs. DuMouchel and Szarfman, who were crucial in the
development of MGPS and its implementation at FDA, respectively, were, in early 2021, leading
an effort to replace that data analysis method with another one. However, it is because of their
expertise and familiarity with the data mining system that Drs. DuMouchel and Szarfman were
uniquely qualified to critique the MGPS methodology, acknowledge its limitations, and offer a
solution to better detect safety signals.46
Records indicate that certain senior FDA officials openly acknowledged their own limited
understanding of data mining, as well as Drs. DuMouchel and Szarfman’s expertise. In a March
2021 email among FDA CBER officials, Dr. Bethany Baer wrote, “I know Ana [Szarfman]
worked to develop the data mining system,” referring to FDA’s current method.47 Dr. Baer
recognized Dr. Szarfman’s “knowledge and experience” with data mining and later added “I
know that she knows a lot more about data mining than I do!”48 In an April 2021 email among
FDA CBER officials discussing Drs. DuMouchel and Szarfman’s data mining analysis, Dr. Craig
Zinderman admitted “I don’t pretend to understand it, but sounds like they are suggesting an
analysis not stratified by year.”49
42 Dr. Szarfman referred to Dr. DuMouchel’s RGPS methodology as “new.” PSI-HHS-000008257443-44; Dr.
DuMouchel co-authored a white paper in 2012 about RGPS. William DuMouchel and Rave Harpaz, Regression-
Adjusted GPS Algorithm (RGPS), Oracle, Nov. 2012, https://docs.oracle.com/health-sciences/empirica-signal-
811/ESIUG/Regression-Adjusted_GPS_Algorithm.pdf; PSI-HHS-000004783470-71; PSI-HHS-000008257443;
PSI-HHS-000008259555, 62; Vaccine Adverse Event Reporting System (VAERS) Standard Operating Procedures
for COVID-19, Centers for Disease Control and Prevention, Jan. 29, 2021,
https://web.archive.org/web/20210319091240/https://www.cdc.gov/vaccinesafety/pdf/VAERS-v2-SOP.pdf at 16-17.
43 PSI-HHS-000008259555, 57.
44 Id. at 61 (emphasis in original).
45 Id.
46 In a March 26, 2021 email to senior FDA officials, Dr. Szarfman left little doubt about her and Dr. DuMouchel’s
expertise in data mining, stating explicitly that she and Dr. DuMouchel “extensively studied the increased value of
RGPS over MGPS for reducing false positives and negative signals.” PSI-HHS-000008257443-44.
47 PSI-HHS-000008257238.
48 PSI-HHS-000008257237.
49 PSI-HHS-000008251980.
10
Throughout 2021, it appears that each time Dr. Szarfman distributed an RGPS-driven
data mining analysis to her colleagues, she also included the results of the MGPS-generated data
mining analysis.50 By doing this, FDA officials could clearly see the alarming statistically
significant safety signals detected by the newer RGPS method compared to the silent and
apparently outdated MGPS method.51
B. Locking Down Access to FDA’s Data Mining Analyses
In September 2021, Dr. Marks informed Dr. Cavazzoni, the then-Director of CDER, that
Dr. Szarfman, who is a CDER employee, “has been asked to cease and desist” conducting her
data analyses.”52 He complained that Dr. Szarfman’s work has “become a major distraction.”53
Dr. Peter Stein, the then-Director of CDER’s Office of New Drugs, reported to Dr. Marks that his
office has “made it clear” to Dr. Szarfman “that she should not be discussing or providing
internal analyses externally, and needs to focus on her assigned work.”54
FDA officials’ decision to prevent Dr. Szarfman from conducting data mining on
COVID-19 vaccines appeared to be part of a larger effort to restrict access and distribution of
FDA’s COVID-19 vaccine data mining analyses. About a month before Dr. Szarfman was
directed to “cease and desist,” Dr. David Menschik, a senior official in FDA CBER who had
been distributing weekly data mining reports to individuals at FDA and CDC, informed a CDC
official, in August 2021, that FDA’s “plan is actually to limit its distribution, largely for data
security reasons.”55 It is unclear what specific “data security reasons” Dr. Menschik referred to
when he informed his CDC colleagues about the need to limit the distribution of the data mining
reports.
Based on records reviewed by the Subcommittee, Dr. Menschik continued to distribute
the weekly data mining reports to a limited number of FDA and CDC officials, but it appears that
the final weekly data mining report for the COVID-19 vaccines was distributed by another FDA
CBER official, Dr. Zinderman, on July 12, 2022.56
In a September 2022 email to his FDA colleagues, FDA CBER official Dr. Narayan Nair
admitted that he was responsible for suggesting to CDC that FDA discontinue its distribution of
50 In a March 26, 2021 email, Dr. Szarfman wrote to her FDA colleagues, “the attached excel comparisons between
RGPS and MGPS were generated by Bill DuMouchel using the VAERS public domain data incorporated into the
Empirica Signal. RGPS is included with the public domain version of Empirica Signal.” PSI-HHS-000008257443;
See e.g., PSI-HHS-000008257443-44 (with attachment); PSI-HHS-000008258306 (with attachment); PSI-HHS-
00008258202-03; PSI-HHS-000002208944-45; PSI-HHS-000004592364 (with attachment).
51 Id.
52 PSI-HHS-000002213753.
53 Id.
54 PSI-HHS-000002213752-53.
55 PSI-HHS-000005524064. See e.g., PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/D4D8B767-A092-455C-9CDE-DC8EC1419B31.
56 PSI-HHS-000001217046 (with attachment); PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/D4D8B767-A092-455C-9CDE-DC8EC1419B31 at 231-234; See
also Dr. Menschik writing to his FDA colleagues on Aug. 5, 2022, “we are no longer routinely sending COVID data
mining to CDC[.]” One FDA official responded, “I was also hoping we would stop doing that at some point!” PSI-
HHS-000008266854.
11
the weekly data mining reports.57 Instead, Dr. Nair informed his colleagues that FDA would
continue its data mining, but would only “notify CDC if we found any datamining [sic] alerts
that were clinical [sic] relevant and required further action.”58 He noted that this method would
save time and “reduce email traffic.”59 Dr. Nair pointed out further, “[s]ince it had been some
time that we had a datamining [sic] alert that required further evaluation, CDC agreed with this
approach.”60
Emails among CDC officials point to another reason behind FDA’s decision to stop
distributing its data mining reports. In November 2022, certain CDC officials recalled how they
used to receive data mining reports from FDA official Dr. Menschik, but no longer do. One
CDC official noted, “I think that because of the FOIAs [Freedom of Information Act
requests] we may have asked FDA to stop sending these weekly data mining outputs .”61
Indeed, in the months and weeks leading up to FDA’s decision to discontinue its
distribution of the weekly data mining reports, there were several public requests made to CDC
and FDA for that information. On May 9, 2022, about two months before the FDA stopped
sending CDC its weekly data mining reports, Children’s Health Defense sent a FOIA request to
CDC for the EB data mining analyses.62 On June 23, 2022, 19 days before FDA sent its final
weekly data mining report, Senator Johnson wrote to then-CDC Director Dr. Rochelle Walensky
requesting the same information.63 Then, on June 30, 2022, 12 days before FDA’s last
distribution, the Informed Consent Action Network submitted a FOIA request to FDA for the EB
data mining records.64
In light of these public and Congressional requests, the timing of FDA’s decision to end
its distribution of its weekly data mining reports seems particularly suspect.65 The notion that
Biden health officials took steps to limit its distribution of COVID-19 vaccine safety data in
order to avoid more internal review or public scrutiny is completely unacceptable.
57 PSI-HHS-000001160286.
58 Id.
59 Id.
60 Id.
61 PSI-HHS-000002480132.
62 FOIA request, May 9, 2022, https://jackanapes.substack.com/api/v1/file/44421d00-9c02-4fc9-9471-
65fcdb24152c.pdf.
63 Letter from Senator Ron Johnson to Rochelle Walensky, Director, Centers for Disease Control and Prevention,
June 23, 2022, https://www.ronjohnson.senate.gov/services/files/9914278B-A73B-4434-8349-91091138E18B.
64 Letter from Aaron Siri, Attorney, Siri Glimstad, et al., to Food and Drug Administration, Oct. 31, 2022,
https://icandecide.org/wp-content/uploads/2023/02/001-Complaint-Exhibits-January-25-2023-IR0802M.pdf at 7.
According to the Informed Consent Action Network, FDA denied the entire FOIA request on Aug. 26, 2022. Id. at
8.
65 The first FOIA production of EB data mining reports did not occur until January 22, 2025, under the Trump
administration. David Wiseman, Signal loss by truancy, masking, and filtering, and underestimation of potential
risks and suspected adverse reactions in the Disproportionality Signal Analyses of VAERS data associated with
COVID-19 pro-vaccines, ResearchGate, Sept. 2025,
https://www.researchgate.net/publication/395382959_Signal_loss_by_truancy_masking_and_filtering_and_underest
imation_of_potential_risks_and_suspected_adverse_reactions_in_the_Disproportionality_Signal_Analyses_of_VAE
RS_data_associated_with_COVID-19_pro at 51, see footnote 41.
12
C. FDA Officials Acknowledge EB Data Mining Limitations and Continue to Fail to Make
Necessary Adjustments
In September 2021, following Dr. Szarfman’s multi-month effort to convince FDA
officials to utilize a better EB data mining algorithm that can adjust for masking, records show
FDA officials acknowledging the masking limitation in the current data mining system.66
However, rather than adhering to Dr. Szarfman’s recommendation to implement a “state of the
art” method that will “lead[] to a lower rate of missed signals,” FDA officials appeared to
continue to use their same data mining method “rather than take an experimental approach.”67
Records show Dr. Menschik working on a draft article in September 2021 about the
safety of mRNA vaccines.68 Dr. Menschik circulated proposed language for that manuscript that
explained EB data mining limitations, including masking. Dr. Menschik wrote:
EB data mining has multiple limitations including that an absence of a
disproportionality alert does not rule out presence of a safety problem.
Additionally, since most reports received during this surveillance period involved
COVID-19 vaccines, disproportionately [sic] scores (which are adjusted by year to
control for time-dependent, potentially confounding, exposure and outcome
variables) can be muted by COVID-19 vaccine reports contributing
substantially to the comparator group, particularly if both mRNA COVID-19
vaccines are associated with the same adverse event.69
In another discussion in September 2021 involving the draft paper, Dr. Menschik also
raised the masking effect, noting:
[I]f the comparison group is enriched with so many mRNA COVID-vaccine
reports, tha[n] it becomes very difficult to exceed the EB05>2 alert threshold
66 PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-A02D-243AA1844D6E at 117.
67 PSI-HHS-000008259561; PSI-HHS-000001136460; PSICOVID_00015642; PSICOVID_00014435.
68 PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-A02D-243AA1844D6E at 117.
69 Id. (emphasis added). As explained in Chairman Johnson’s March 23, 2026 letter to HHS, the Subcommittee
uncovered several emails regarding the evolution of this excerpt in the manuscript. Similar language referencing EB
data mining limitations appeared into two preprint versions of the manuscript, one dated Oct. 27, 2021, and one
dated Oct. 28, 2021. The final version of the paper appeared in the Lancet in March 2022 without the reference to
EB data mining limitations. See Letter from Ron Johnson, Chairman, Permanent Subcomm. on Investigations to
Robert F. Kennedy, Jr., Sec., Dep’t of Health and Human Services, Mar. 23, 2026,
https://www.ronjohnson.senate.gov/services/files/CA500350-195E-472C-9F26-BE93B290B9D9; See Hannah
Rosenblum, et al., Safety Monitoring of mRNA Vaccines Administered During the Initial 6 Months of the U.S.
COVID-19 Vaccination Program: Reports to Vaccine Adverse Events Reporting System (VAERS) and v-safe,
MedRxiv, Oct. 27, 2021, https://www.medrxiv.org/content/10.1101/2021.10.26.21265261v1.full.pdf at 13; Hannah
Rosenblum, et al., Safety Monitoring of mRNA Vaccines Administered During the Initial 6 Months of the U.S.
COVID-19 Vaccination Program: Reports to Vaccine Adverse Events Reporting System (VAERS) and v-safe,
MedRxiv, Oct. 28, 2021, https://www.medrxiv.org/content/10.1101/2021.10.26.21265261v2.full.pdf at 13; Hannah
Rosenblum et al., Safety of mRNA vaccines administered during the initial 6 months of the US COVID-19
vaccination programme: an observational study of reports to the Vaccine Adverse Event Reporting System and v-
safe, The Lancet, https://pmc.ncbi.nlm.nih.gov/articles/PMC8901181/.
13
for an adverse event that may be associated with mRNA vaccines – thus data
mining has blind spots and this is why it’s so good to have so many complimentary
vaccine safety surveillance systems (e.g., VSD [Vaccine Safety Datalink]) that can
cover different blind spots of other systems…70
In November 2022, Dr. Menschik emailed Dr. Nair and referenced the masking limitation
which could lead to misinterpretations of the data mining.71 In October 2023, Dr. Nair wrote a
comprehensive email to federal health officials including, Drs. John Su and Tom Shimabukuro,
both from CDC, about the masking effect in data mining. Dr. Nair explained:
As you know, data mining has all the limitations of passive surveillance as
well as others. However, during the COVID vaccine era there is an additional
limitation. Since most reports received involve COVID-19 vaccines,
disproportionately [sic] scores (which are adjusted by year to control for time-
dependent, potentially confounding, exposure and outcome variables) can be driven
towards the null by COVID-19 vaccine reports contributing substantially to the
comparator group.72
Dr. Nair added, “We were aware of this limitation before and during the pandemic.
There are many data mining tools and there was some discussion about utilizing a novel tool
to adjust for this. However, we thought it would be problematic to use a brand new, possibly
unvalidated tool in the context of an EUA.”73 It appears likely that Dr. Nair’s reference to a
“novel tool to adjust for” masking was Drs. Szarfman’s and DuMouchel’s proposed newer data
mining algorithm. Although Dr. Nair claimed the tool was “possibly unvalidated,” he failed to
mention its efficacy in detecting statistically significant safety signals which, one would think,
would be invaluable especially if it is being used to monitor the effects of a novel vaccine that
just received an EUA.
D. Top FDA CBER Officials Eventually Acknowledge the Effectiveness of the New Data
Mining Model Compared to the Current Methodology
In September 2024, CBER officials Drs. Baer and Menschik discussed FDA’s current
data mining methodology, MGPS, and RGPS, the methodology that Dr. Szarfman had lobbied
for years earlier.74 Dr. Baer noted that when she compared results of the RGPS and MGPS
models for the Gardasil vaccine, she found that the RGPS model yielded “significantly higher”
numbers than the MGPS model for certain adverse events.75 She also found that the RGPS
numbers were higher than MGPS numbers for certain adverse events associated with the Pfizer
70 PSI’s March 25, 2026 document release on EB data mining:
https://www.ronjohnson.senate.gov/services/files/7DD643F9-BF66-4923-B791-7766870C8856 at 120 (ellipses in
original, emphasis added).
71 PSICOVID_00015642.
72 PSI-HHS-000001136460.
73 Id. Dr. DuMouchel co-authored a paper in 2012 about the RGPS data mining method. William DuMouchel and
Rave Harpaz, Regression-Adjusted GPS Algorithm (RGPS), Oracle, Nov. 2012, https://docs.oracle.com/health-
sciences/empirica-signal-811/ESIUG/Regression-Adjusted_GPS_Algorithm.pdf.
74 PSI-HHS-000008253424.
75 Id. It is unclear what adverse events Dr. Baer reviewed that were associated with the Gardasil vaccine.
14
COVID-19 bivalent vaccine.76 Dr. Menschik appeared to concur with Dr. Baer’s finding and
wrote that “in general the [RGPS model] appears way more sensitive in that its scores are
generally higher than corresponding [MGPS model] scores when sampling different [adverse
events].”77
In the same email, Dr. Baer, who mentioned Dr. Szarfman’s 2022 paper about the
masking limitation in data mining, appeared to acknowledge that the lower numbers from the
MGPS model were a result of masking.78 She wrote, “I understand the theory behind masking
and trying to adjust for it, but I feel that comprehending the details of the approach and,
importantly, which approach is ‘better,’ is beyond my training and experience. I think someone
with more data mining expertise would have to be involved in that decision.”79
Drs. Menschik and Baer’s September 2024 email exchange acknowledging the
effectiveness of the RGPS model compared to the MGPS model and referencing the need for a
data mining expert is beyond ironic. Since 2021, because of Drs. Szarfman and DuMouchel’s
efforts, top FDA officials, including Drs. Marks and Menschik, were well aware of the
effectiveness of the RGPS model as a means to adjust for masking and better detect statistically
significant safety signals. Dr. Baer herself acknowledged in March 2021 that Dr. Szarfman had
“knowledge and experience” with data mining and even admitted at that time, “I know that [Dr.
Szarfman] knows a lot more about data mining than I do!”80 FDA officials had ample
opportunity to involve Drs. Szarfman and DuMouchel—both experts in data mining—in their
safety surveillance efforts, but instead, cast them and their analyses aside.
# # # #
The enclosed timeline will detail Dr. Szarfman’s attempts to inform her colleagues at the
FDA about increases in statistically significant safety signals associated with the COVID-19
vaccines. The timeline will also show how FDA officials reacted to Dr. Szarfman’s findings,
their growing concern about her access to the safety data, and their failure to adjust the FDA’s
data mining analysis to account for the significant flaws in their methodology that masked
dozens of statistically significant adverse events for the COVID-19 vaccines.
76 Id. It is also unclear what adverse events Dr. Baer reviewed that were associated with the Pfizer COVID-19
bivalent vaccine.
77 Id.
78 Id.
79 Id.
80 PSI-HHS-000008257237-38.
15
II. Timeline – Biden Health Officials Ignore Safety System that Can Better
Detect Signals for COVID-19 Vaccine Adverse Events
Nov. 18, 2020: Manette Niu, an official at the Food and Drug Administration’s (“FDA”)
Center for Biologics Evaluation and Research (“CBER”), Office of Biostatistics and
Pharmacovigilance (“OBPV”), emails Steven Anderson, the Director of that office at that
time, to inform him that she has a meeting the next day with Ana Szarfman, a medical
officer and safety data mining developer who works in FDA’s Center for Drug Evaluation
and Research (“CDER”), Office of Cardiology, Hematology, Endocrinology, and
Nephrology (“OCHEN”), “to discuss [Szarfman’s] proposal on instituting new methods
to improve data mining.”81 It is unclear what specific “new methods” Szarfman wanted
to discuss and whether this meeting occurred.
Dec. 11, 2020: FDA issues an Emergency Use Authorization (“EUA”) for the Pfizer
COVID-19 mRNA vaccine for individuals ages 16 and older.82
Dec. 18, 2020: FDA issues an EUA for the Moderna COVID-19 mRNA vaccine for
individuals ages 18 and older.83
Dec. 23, 2020: Szarfman organizes a virtual meeting to discuss proposals for a new
safety data mining methodology designed by Oracle Chief Statistician William
DuMouchel, the architect of FDA’s existing data mining system.84 Meeting invitees
include Richard Forshee, Associate Director for Research at CBER OBPV/Acting Deputy
Director for CBER OBPV; Narayan Nair, a CBER OBPV official; and Norman
Stockbridge, an official at CDER OCHEN.85 Szarfman notes in her introductory remarks
for the meeting that among other problems with current data mining practices “safety
signals may remain hidden.”86
Jan. 8, 2021: As FDA officials discussed the use of empirical Bayesian (“EB”) data
mining to detect safety signals for adverse events associated with COVID-19 vaccines,
David Menschik and Bethany Baer, two officials at CBER OBPV, communicate about “a
few classic data mining references” written by Szarfman and DuMouchel.87 Baer notes
81 PSI-HHS-000004585100. Until 2022, The Office of Biostatistics and Pharmacovigilance was named the Office of
Biostatistics and Epidemiology. For purposes of this document, the office will be referred to by its current name.
See National Science Foundation, NI Protocol, FDA CBER Reorg & More—March ’22 Pharma News, Mar. 1,
2022, available at https://www.nsf.org/news/march-pharma-news.
82 COVID-19 Timeline, Centers for Disease Control and Prevention,
https://www.cdc.gov/museum/timeline/covid19.html.
83 Id.
84 PSICOVID_00014194.
85 PSICOVID_00014194; PSI-HHS-000002185990-91. Szarfman’s Dec. 23, 2020 email thanking Forshee for
attending the presentation notes that she has attached an updated version of the slides. One such update appears to
be a slide including DuMouchel’s background as the inventor of MGPS. See PSI-HHS-000002293255.
86 PSICOVID_00014198.
87 PSI-HHS-000008254390.
16
that FDA’s current EB data mining system uses the Multi-item Gamma Poisson Shrinker
(“MGPS”), which is the algorithm DuMouchel invented.88
Jan. 29, 2021: The Centers for Disease Control and Prevention (“CDC”) releases its
Standard Operating Procedures (“SOP”) for the Vaccine Adverse Event Reporting System
(“VAERS”).89 The SOP identifies the “analyses for COVID-19 vaccine safety signals”
noting that FDA will be responsible for “using empirical Bayesian data mining to identify
[adverse events] reported more frequently than expected following vaccination with
COVID-19 vaccines.”90
Feb. 14, 2021: Szarfman raises concerns to CBER Director Peter Marks regarding
COVID-19 vaccine safety monitoring issues with the FDA’s current practices and asks “if
you want Bill DuMouchel and I to discuss our proposal for more effective monitoring,
including the need to use an updated algorithm by DuMouchel for data mining
spontaneous reports at [FDA.]”91 Marks arranges a time to meet with her in the
following weeks.92
Feb. 27, 2021: FDA approves an EUA for Johnson & Johnson’s COVID-19 vaccine for
all people ages 18 years and older.93
March 1, 2021:
o Szarfman meets with CBER Director Peter Marks, as well as Forshee, Anderson,
and others at CBER, to discuss confounding issues with MGPS, particularly data
masking, which is a known limitation with MGPS that can lead to missed or
reduced signals in data mining.94 Szarfman also showcases “the superior
performance” of DuMouchel’s Regression-Adjusted Gamma Poisson Shrinker
(“RGPS”), an “updated algorithm” for data mining that DuMouchel first outlined
in a 2012 white paper.95 Szarfman states that “[t]he MGPS data mining method
currently in use at [FDA]” is “not the state of the art.”96 In contrast, she
emphasizes, “RGPS is the state of the art,” noting that the new method can
“better adjust for both, masking (false negatives) and confounding (false
positives).”97
88 Id.; PSI-HHS-000002293255.
89 Vaccine Adverse Event Reporting System (VAERS) Standard Operating Procedures for COVID-19, Centers for
Disease Control and Prevention, Jan. 29, 2021,
https://web.archive.org/web/20210319091240/https:/www.cdc.gov/vaccinesafety/pdf/VAERS-v2-SOP.pdf.
90 Id. at 16-17.
91 PSI-HHS-000002213369.
92 Id.
93 COVID-19 Timeline, Centers for Disease Control and Prevention,
https://www.cdc.gov/museum/timeline/covid19.html.
94 PSI-HHS-000004783471-72; PSI-HHS-000002134656.
95 PSI-HHS-000004783471-72; PSI-HHS-000008259555; William DuMouchel and Rave Harpaz, Regression-
Adjusted GPS Algorithm (RGPS), Oracle, Nov. 2012, https://docs.oracle.com/health-sciences/empirica-signal-
811/ESIUG/Regression-Adjusted_GPS_Algorithm.pdf.
96 PSI-HHS-000008259561.
97 Id. (emphasis in original); PSI-HHS-000002134652 (emphasis added).
17
o Marks thanks Szarfman “for taking the time to go over everything so carefully
with us.”98 He writes, “[w]e will work through the issues that you presented.”99
Afterwards, Marks schedules a meeting with Forshee and Anderson to discuss
“how to proceed” apparently on the issues raised by Szarfman and DuMouchel’s
updated data mining algorithm.100 It is unclear whether this meeting occurred.
March 10, 2021: Szarfman confirms that she has access to VAERS data through Oracle’s
Empirica, a tool for detecting and analyzing vaccine safety signals.101 It appears other
CBER officials, including Anderson, were aware of and did not object to Szarfman’s
access, even though she works in CDER.102
March 11, 2021: Craig Zinderman, a CBER official, affirms to Baer that he has no issue
granting Szarfman access to VAERS data through Empirica.103 Baer responds writing
“we will leave Ana’s account as is with a CBER login.”104
March 12, 2021:
o A day later, Baer appears to take issue with Szarfman having access to a database
controlled by CBER, an office that Szarfman is not a part of, saying, on calls with
the contractor supporting Empirica and FDA’s data mining, Commonwealth
Informatics (“Commonwealth”) Szarfman “has twice now expressed interest in
COVID vaccine data mining and made some broad statements that I don’t think
[the Division of Pharmacovigilance] would agree with[.]”105 Baer also questions
who at CBER will be coordinating with Szarfman and interpreting her results.106
However, Baer appears to recognize Szarfman’s expertise in data mining and
writes, “I know Ana [Szarfman] worked to develop the data mining system
and this might be a special circumstance due to her knowledge and
experience.”107
o Zinderman, replies that “[r]efusing her access just for vaccines seems a little
disingenuous” but that it “[s]eems reasonable to try to understand why [Szarfman]
wants to use VAERS data instead of [CDER’s] data, and to caution her that while
its [sic] fine for her to do methodological work, we aren’t interested in
additional data mining studies of COVID data outside of CBER’s usual
processes.”108
98 PSI-HHS-000004783470-71.
99 Id.
100 PSI-HHS-000004584613.
101 PSI-HHS-000008257240-41.
102 PSI-HHS-000008257239.
103 PSI-HHS-000008257238-39.
104 PSI-HHS-000008257238.
105 Id. At the time of Baer’s email the Division of Pharmacovigilance was known as the Division of Epidemiology.
106 Id.
107 Id. (emphasis added).
108 PSI-HHS-000008257237-38 (emphasis added).
18
March 14, 2021: Responding to Zinderman, Baer acknowledges that “[Szarfman] knows
a lot more about data mining than I do[.]”109 However, Baer reiterates her earlier
concerns about Szarfman’s apparent comments she made during a regularly scheduled
call between officials at CBER, CDER, and Commonwealth.110 Baer characterizes
Szarfman’s comments as atypical and speculates that Szarfman may have been “just
brainstorming and theorizing” which was “unusual compared to the typical topics
covered on the call.”111
March 16, 2021: Niu writes to Baer that she and Zinderman have decided to schedule a
meeting with Szarfman to discuss her “VAERS objectives/rationale.”112
March 17-18, 2021:
o Menschik emails Brian Hendrix, a contractor at Commonwealth, and Baer to
request a “‘special project’ run” of MGPS that will apparently adjust for masking
with COVID-19 vaccines, though he does not use that specific term.113 He notes
that he observed a “muting trend” for adverse events, which he attributes to the
volume of COVID-19 vaccine reports.114 Menschik specifies that the viewers of
this potential data analysis should be limited to himself, Baer, and the
contractor.115
o The “‘special run’” was completed, but the results do not appear in the records
reviewed to date.116
March 26, 2021:
o After speaking with Zinderman, Niu, and Baer earlier that day, Szarfman sends
them, as well as Menschik and Stockbridge, an earlier email and spreadsheet from
DuMouchel that includes his RGPS analysis showing “49 examples of extreme
masking,” with over twenty of those examples of adverse events now showing a
statistically significant safety signal when adjusted for masking.117 Some of the
statistically significant safety signals for adverse events associated with the
COVID-19 vaccines that were not previously detected through MGPS include
sudden cardiac death, Bell’s palsy, and pulmonary infarction.118 Szarfman also
attaches her March 1 presentation she gave to Marks and other FDA officials.119
o Szarfman notes that the “comparisons between RGPS and MGPS were generated
by Bill DuMouchel using the VAERS public domain data incorporated into
Empirica Signal.” Szarfman explains that she and DuMouchel “extensively
studied the increased value of RGPS over MGPS for reducing false positives and
109 PSI-HHS-000008257237.
110 Id.
111 Id.
112 Id.
113 PSICOVID_00017214.
114 Id.
115 Id.
116 PSICOVID_00017210.
117 PSI-HHS-000008263190-91 (emphasis added).
118 PSI-HHS-000008257443-44 (with attachment).
119 Id.
19
negative signals.”120
March 30, 2021: Menschik circulates a list of adverse events with safety signals
identified by FDA’s data mining methodology to FDA senior officials, as well as several
officials at CDC, including John Su and Tom Shimabukuro.121 Menschik also attaches a
draft presentation detailing limitations with the FDA’s current MGPS data mining
system.122 Some of the limitations highlighted in the draft presentation, such as
“confounding” and “false alerts from statistical interaction,” appear to be similar to the
concerns Szarfman previously raised regarding the effectiveness of the FDA’s MGPS
algorithm.123 It is unclear how the senior CDC and FDA officials responded to the draft
presentation.
April 10, 2021:
o DuMouchel sends additional data to Szarfman and others that appears to analyze
VAERS reports while adjusting for the fact that “99% of all reports” for the year
2021 are for the COVID-19 vaccines. 124 DuMouchel’s analysis appears to adjust
for the masking effect. 125 He notes, “[o]nly if you mix in more non-covid reports
within each stratum can you get enough diversity to allow larger
disproportionalities.” 126
o Szarfman forwards the email to Niu and other FDA officials, flagging the
“[i]mportant analysis by DuMouchel,” and writes, “I think that we need to invite
[DuMouchel] to talk with us about the effect of adjustment factors, given the data,
so we can all learn from his knowledge.”127
April 12-13, 2021: Two days after Szarfman flags DuMouchel’s “[i]mportant analysis,”
Niu replies to her in a separate chain asking, “Does this effect the data mining results we
are receiving in 2021?”128 Szarfman responds the next day, saying, “Exactly. As
DuMouchel pinpointed, there is a need to extend the stratification brackets by the fact
that 99% of the results for FY2021 are for COVID-19 vaccines this indeed affects the
results.”129
April 14, 2021: Niu forwards the emails containing DuMouchel’s writeup of his analysis
and Szarfman’s comments to Zinderman (removing Szarfman from the email chain), but
it does not appear Niu shares DuMouchel’s data mining results at this time.130 Zinderman
in turn sends DuMouchel’s writeup and Szarfman’s comments to Menschik and Baer with
120 Id.
121 PSI-HHS-000001187885 (with attachment); PSI-HHS-000001187886-88.
122 Id.
123 PSI-HHS-000001187888.
124 PSI-HHS-000008251981-82; PSI-HHS-000008258306-07 (with attachment).
125 Id.
126 PSI-HHS-000008251982.
127 PSI-HHS-000008251981-82.
128 PSI-HHS-000008251981.
129 PSI-HHS-000008251980-81.
130 Id.
20
Niu copied.131 Zinderman questions whether DuMouchel’s apparent adjustment for
masking “[m]ight be worth considering[.]”132 He adds, “I don’t pretend to understand it,
but sounds like they are suggesting an analysis not stratified by year.”133
April 15, 2021:
o The next day, Baer responds, stating, “I think we should welcome any expert
input,” acknowledging that she and Menschik “are concerned about the effect of
so many COVID reports on the standard system [FDA] use[s].”134 She asks if
there is any way to have DuMouchel more involved in FDA’s “data mining
process and interpretation during this unprecedented reporting time?”135
o Niu answers stating she will send them DuMouchel’s data mining results and that
“[t]he best person to ask would be Ana [Szarfman] as she has close ties with Bill
DuMouchel.”136
o A few minutes later, Menschik removes Niu from the email chain and replies only
to Zinderman and Baer, and appears to push back on immediately speaking with
Szarfman, saying, “Before we potentially reach out to Ana [Szarfman], we
should meet internally – many considerations not suited to email…”137 It is
unclear if Menschik, Zinderman, and Baer met.
April 17-19, 2021:
o Szarfman forwards Niu an analysis that DuMouchel compiled the day before
using his new data mining method (RGPS), which revealed six statistically
significant safety signals for adverse events, which were not previously detected
by FDA’s current system.138 DuMouchel’s analysis uncovered statistically
significant safety signals, currently masked by MGPS, including Bell’s palsy
associated with the Pfizer vaccine and thrombosis associated with the Moderna
vaccine.139
o Two days later, Niu forwards Szarfman’s email containing DuMouchel’s analysis
to Zinderman, Baer, and Menschik.140
April 22, 2021:
o After receiving Szarfman’s distribution of DuMouchel’s analysis, Baer informs
Menschik, Zinderman, and Niu that during a call that day with Commonwealth
officials, Szarfman “offered to show individuals the interesting VAERS analysis
131 Id.
132 PSI-HHS-000008251980.
133 Id.
134 Id.
135 Id.
136 PSI-HHS-000008251979-80.
137 PSI-HHS-000008251979 (emphasis added) (ellipses in original).
138 PSI-HHS-000008258202-03. Records show that FDA’s threshold for determining a statistically significant safety
signal was when the lower bound of the reporting estimate (EB05) exceeded 2.0. See PSI’s March 25, 2026
document release on EB data mining https://www.ronjohnson.senate.gov/services/files/F92FBB0A-CCFD-412A-
A02D-243AA1844D6E at 166.
139 PSI-HHS-000008258203.
140 PSI-HHS-000008258202.
21
she has been doing with Manette. A couple of the Commonwealth folks
expressed interest in meeting with her to see it.”141
o Niu responds, distancing herself from Szarfman, writing, “I have not been
working with Ana directly, although she has sent me data mining runs that I’ve
forwarded to this group. I will speak to her about this.”142
April 25-26, 2021:
o Szarfman resends DuMouchel’s analysis to Niu and includes several additional
FDA officials.143 In the email, Szarfman reiterates that she is sending all of
DuMouchel’s analysis to Niu.144 Szarfman further writes regarding how
DuMouchel’s new method (RGPS) “is better at removing false positives and
negatives than MGPS. Note the safety signals for cardiac events with the
Pfizer and Moderna vaccines, now in the news, that are better identified by
RGPS than by MGPS.”145
o Niu separately asks Zinderman for advice on how to respond to Szarfman who
had apparently described Niu as her collaborator on Szarfman’s data mining work
to FDA’s data mining contractors.146 In response, Zinderman sends Niu draft
language to respond to Szarfman, and tells Niu that asking Szarfman to stop
sending data mining results “would create a bigger problem for us I think.”147
o Niu, responds to Szarfman’s earlier email and writes, “[w]hile we are aware that
CDER is using the vaccine data to explore new calculations and various
deviations of analysis parameters in disproportionality analysis, I haven’t been,
and are unable to, work as a collaborator with you on this project due to our
higher priority work, and because this sort of statistical development work falls
outside of my area of expertise.”148 Niu’s email to Szarfman contains similar
language to the draft response Zinderman sent Niu earlier that day.149
o Szarfman responds to Niu that she understands but she will keep her up to date
with her findings.150
April 29, 2021:
o Szarfman shares with Niu and other FDA officials an analysis of EB data mining
issues unrelated to DuMouchel’s new data mining method.151 Niu forwards
Szarfman’s analysis to Zinderman, Baer, and Menschik.152 In response, Menschik
asks, “[d]id you request this or anything else (COVID vaccine data mining
141 Id.
142 Id.
143 PSI-HHS-000008258271.
144 Id.
145 Id. (emphasis added).
146 PSI HHS-000008260150.
147 Id.
148 PSI-HHS-000008258271.
149 PSI HHS-000008260150.
150 PSI-HHS-000008258271.
151 PSI-HHS-000008258153-54; PSICOVID_00017031; PSICOVID_00017032-33; PSICOVID_00017034-81.
152 PSI-HHS-000008258153.
22
related) from Ana and/or are you working with Ana on any data mining projects?
(if so, please specify)[.]”153
o Niu responds to Menschik in two separate emails, writing first, “[n]o, I haven’t
requested anything from Ana. I am only passively passing on her data mining runs
when she sends them to me.”154 She then quickly sends the second email stating
that she is “not working on anything with [Szarfman].”155
o The same day, Niu also sends Menschik her previous conversation with Szarfman,
in which Niu told Szarfman that she cannot collaborate with Szarfman.156
May 6, 2021: Szarfman circulates an analysis run by DuMouchel to Norman Stockbridge
and another CDER official, showing that MGPS had undervalued the safety signals for
acute myocardial infarction for both the Pfizer and Moderna COVID-19 vaccines.157
DuMouchel’s RGPS method appeared to yield much higher signals compared to FDA’s
MGPS method.158
May 7, 2021:
o Following Szarfman’s multi-month efforts advocating for FDA to change its data
mining system in order to unmask statistically significant safety signals associated
with the COVID-19 vaccines, Menschik, Nair, and Zinderman draft a lengthy
email to Szarfman.159
o The final draft of the email, which Zinderman eventually sends to Szarfman that
same day, directs her to “please hold off on creating and sending data mining
reports and analyses using COVID-19 vaccine [adverse event] data.”160
o In the email to Szarfman, the FDA officials attempt to downplay her findings
writing that, “we are already screening and reviewing reports, such as [acute
myocardial infarction], [thrombocytopenia syndrome], Thromboembolic events,
and other forms of coagulopathy.”161 The email concludes by reiterating the
directive to Szarfman to “hold off” on continuing her efforts.162
o Szarfman replies to Zinderman that from now on, “I will only deliver analyses
when I am specifically requested to do so[.]”163 She reiterates, however, “[w]e
are testing a new data mining methodology, and given the circumstances, it will
be good for all to understand its performance with such important data. This is a
method that also strongly reduces confounding , so it may be helpful in certain
future circumstances.”164
153 Id.
154 Id.
155 PSI-HHS-000008258190.
156 PSI-HHS-000008258271.
157 PSI-HHS-000002208944-45.
158 Id. (using the RGPS methodology, ER05 notification signals were significantly higher than the previously
running EB05 of MGPS).
159 PSICOVID_00017246; PSI-HHS-000008251530; PSI-HHS-000008251912-13; PSI-HHS-000001195617-19;
PSI_HHS-000008253450-51; PSI-HHS-000001175745-47; PSI-HHS-000001148712-14.
160 PSICOVID_00017246-47.
161 Id.
162 Id.
163 PSICOVID_00017245.
164 Id. (emphasis added).
23
o Zinderman forwards Szarfman’s response to Menschik and Niu.165
May 10, 2021: FDA expands EUA for Pfizer’s COVID-19 mRNA vaccine for ages 12-
15.166
June 4, 2021: Szarfman receives clearance from Stockbridge to publish a paper she co-
authored with DuMouchel and others at Oracle, entitled “Masking Associated with Early
COVID-19 Vaccine Safety Surveillance.”167 Szarfman encloses a draft version of her
paper that “demonstrates the potential utility of a new signal detection methodology
called RGPS that can address masking and confounding effects that cannot be properly
controlled by conventional signaling methodologies.”168
June 11, 2021:
o Following apparent conversations between the CDC and FDA regarding
“myocardial events and the lack of signals in VAERS[,]” Szarfman writes to her
FDA colleagues, resending DuMouchel’s analysis that she initially sent one
month prior that uncovered statistically significant safety signals for acute
myocardial infarction.169 Referring to the FDA and CDC’s discussion on
myocardial events, Szarfman writes, “I am not astonished that MGPS was
unable to detect these signals.”170 Once again, she reiterates the strength of her
and DuMouchel’s new data mining system, noting that one month ago, “we
documented that RGPS signals [acute myocardial infarction].”171
o Szarfman eventually alerts Forshee, a high-ranking CBER official, about her
findings relating to statistically significant safety signals for acute myocardial
infarction and invites him to collaborate with her on her upcoming paper about
masking.172 It is unclear whether Forshee responds.173
July 9, 2021: Szarfman calls Forshee to discuss increased risk of mortality following
COVID-19 vaccination.174
July 12, 2021: Following her July 9 phone call, Szarfman writes Forshee, “As we talked
over the phone, I became aware last Fri[day] that scientists from Cornell are concerned of
an increased mortality signal with the COVID-19 vaccines.”175 Szarfman encloses
DuMouchel’s earlier RGPS analysis, showing statistically significant safety signals
for “death and sudden death” associated with the Pfizer and Moderna COVID-19
165 Id.
166 COVID-19 Timeline, Centers for Disease Control and Prevention,
https://www.cdc.gov/museum/timeline/covid19.html.
167 PSI-HHS-000001640809; PSI-HHS-000001622854-55.
168 PSI-HHS-000001622854; PSI-HHS-000001639129-34 at 33.
169 PSI-HHS-000002208944-45.
170 PSI-HHS-000002208944.
171 Id.
172 Id.
173 Id.
174 PSI-HHS-000002199145-46; PSI-HHS-000004592364-65 (with attachment).
175 PSI-HHS-000004592364-65 (with attachment).
24
vaccines.176 She writes that the method DuMouchel used “automatically unmask[s]
signals that remain hidden by other data mining methodologies, including MGPS[,]”
FDA’s current data mining methodology.177
July 13, 2021: Forshee documents his contact with Szarfman and forwards her email and
DuMouchel’s analysis to Marks, noting that, “she and Bill DuMouchel had found an
increased risk of mortality following COVID-19 vaccination using data mining
methods.”178 Forshee writes that he is “very concerned” with how Szarfman and
DuMouchel may be interpreting their findings.179 Marks thanks Forshee for
“documenting this” and tells Forshee that he “will follow up appropriately.”180
Aug. 11, 2021: Menschik informs CDC officials that FDA will limit its distribution of its
data mining reports “largely for data security reasons.”181
Aug. 12, 2021: Szarfman attends a call with Commonwealth, the contractor that runs
FDA’s data mining system, and apparently raises concerns about the current
methodology.182 Szarfman reportedly tells the contractor that “‘the 20-year-old MGPS
model could potentially mask signals.’”183 Baer, who participated on the call, reports
Szarfman’s comments to Zinderman and Menschik and informs them that she told
Szarfman that Zinderman and Menschik are aware of her considerations.184
Aug. 23, 2021: FDA approves the Biologics License Application (“BLA”) for Pfizer’s
Comirnaty COVID-19 vaccine for ages 16 and older.185
Sept. 3, 2021:
o Commonwealth employees, Brian Hendrix and James Sydnor, follow up with
Szarfman from an earlier conversation, informing her that for questions regarding
data mining runs, she needs to speak with Menschik or Baer directly.186 Szarfman
subsequently clarifies her masking concerns with the contractors, writing,
“[t]herefore the background will only be for covid-19 vaccines, instead of for
other vaccines. Therefore, masking covid-19 vaccine signals that are common
with these vaccines, but not common across other types of vaccines.”187
176 Id. (emphasis added).
177 Id. That same day, Forshee forwards Szarfman’s email and DuMouchel’s analysis to Anderson. PSI-HHS-
000004590546-47 (with attachment).
178 PSI-HHS-000004588545-46 (with attachment); PSI-HHS-000004592364-65 (with attachment).
179 PSI-HHS-000004588545-46.
180 PSI-HHS-000002199145.
181 PSI-HHS-000005524064.
182 PSICOVID_00017545.
183 Id.
184 Id.
185 BLA approval, Food and Drug Admin., Aug. 23, 2021,
https://www.fda.gov/media/151710/download?attachment; COVID-19 Timeline, Centers for Disease Control and
Prevention, https://www.cdc.gov/museum/timeline/covid19.html.
186 PSI-HHS-000008254473-74.
187 PSI-HHS-000008254472 (emphasis added).
25
o Hendrix responds to Szarfman, adding Menschik and Baer to the email chain for
their awareness.188 Szarfman replies directly to Menschik, reiterating the
advantages of RGPS compared to MGPS.189 She explicitly writes, “RGPS is
much, much better at unmasking signals than MGPS. It automatically
identifies and corrects for confounders. This is an important function to have,
given the pandemic situation.”190
Sept. 8, 2021: Menschik and Zinderman draft a response to Szarfman’s previous email,
and Menschik sends it to her.191 In his email to Szarfman, Menschik underscores that
VAERS data mining is CBER’s responsibility and must be only conducted through that
office (Szarfman is a CDER employee, not a CBER employee).192 However, Menschik
appears to acknowledge a limited issue with MGPS, separate from masking, that he plans
to review further within CBER.193 He also writes, “[a]ny further discussion on VAERS
data mining methods/findings outside my chain of command will have to be offline
and in general terms, as well as without reference to any specific VAERS [adverse
events].”194
Sept. 14, 2021:
o Anderson and Nair draft an email regarding Szarfman’s data mining activities that
Anderson eventually send to Marks.195
o In the email, Anderson takes issue with Szarfman’s efforts and access to CBER
data “given that she is a CDER employee.”196 He notes, “we have concerns about
her communicating data mining findings using CBER VAERS data to CBER and
non-CBER personnel.”197 Anderson requests that Szarfman “refrain[s] from using
her FDA email or communicating data mining findings using CBER VAERS
data[.]”198
o Anderson appears to acknowledge the masking issue Szarfman has previously
raised.199 Rather than take immediate steps to adjust FDA’s system to address
Szarfman’s concerns, Anderson suggests taking a “retrospective approach” to
determine if safety signals were missed only after receiving reports of adverse
events through active surveillance systems.200 Given that FDA failed to detect the
188 Id.
189 PSI-HHS-000008254471.
190 Id. (emphasis added).
191 PSI-HHS-000008253098; PSI-HHS-000008253959; PSI-HHS-000008252951; PSI-HHS-000008253958; PSI-
HHS-000008267103; PSI-HHS-000008254470.
192 PSI-HHS-000008254470.
193 Id.
194 Id. (emphasis added).
195 PSICOVID_00014054-55; PSI-HHS-000002213753-54.
196 PSI-HHS-000002213753-54.
197 Id.
198 Id.
199 Id.
200 Id.
26
safety signal for myocarditis months earlier, it is unclear what further evidence
was needed to trigger a reevaluation of FDA’s current data mining system.201
Sept. 15, 2021:
o Marks forwards Anderson’s write-up concerning Szarfman’s data mining efforts
to CDER Director Patrizia Cavazzoni, to whom Szarfman reports, and writes, “I
really am sorry to bother you with this, but issue is [sic] become a major
distraction. . . . Ana Szarfman[] has decided on her own to do vaccine analysis
using VAERS as part of her work at FDA. She is, however, not doing this in
collaboration with our CBER statisticians, and quite to the contrary, has been
asked to cease and desist, because the strategy that she is using could create
erroneous conflicts that feed in to anti-vaccination rhetoric.”202
o CDER Office of New Drugs Director Peter Stein replies to Marks, “[T]hanks for
flagging this – we’ve made it clear to her that she should not be discussing or
providing internal analyses externally, and needs to focus on her assigned work.”
He closes his email, “Hopefully, you won’t have further surprises….”203
Sept. 22, 2021:
o Menschik writes to John Su, a CDC official and member of the Vaccine Safety
Team and Vaccine Safety Technical Work Group, and appears to identify
limitations within FDA’s data mining system.204 Menschik describes the masking
effect, the exact data limitation that Szarfman had been raising for months at
FDA.205 Although Menschik does not credit Szarfman, he echoes her concerns in
draft language Menschik shares with Su for an article reviewing the safety of the
COVID-19 vaccines.206
o Menschik’s draft language notes that FDA’s current data mining system “has
multiple limitations,” including that, “disproportionately scores . . . can be
muted by COVID-19 vaccine reports contributing substantially to the
comparator group, particularly if both mRNA COVID-19 vaccines are
associated with the same adverse event.”207 Menschik tells Su that he is “happy”
to discuss further by phone.208
o Su acknowledges the concern in a response email, noting that signal detection
with VAERS data “has always been tricky business” and that those who work
with VAERS frequently are “mindful of those limitations[.]”209
201 See Chairman Ron Johnson, Failure to Warn: How Federal Health Agencies Downplayed the Risk of Myocarditis
and Other Adverse Events Following COVID-19 Vaccination, Permanent Subcomm. on Investigations, May 21,
2025, https://www.hsgac.senate.gov/wp-content/uploads/2025.05.21-PSI-Majority-Staff-Interim-Report-Failure-
toWarn.pdf.
202 PSI-HHS-000002213753 (emphasis added).
203 PSI-HHS-000002213752-53 (ellipses in original).
204 PSI-HHS-000008268909.
205 Id.
206 Id.
207 Id. (emphasis added).
208 Id.
209 Id.
27
Oct. 29, 2021: FDA expands EUA for the Pfizer COVID-19 vaccine for children ages 5-
11.210
Jan. 3, 2022: FDA grants EUA for Pfizer COVID-19 boosters for children ages 12-15
and a third primary series dose for children ages 5-11 who are immunocompromised.211
Jan. 31, 2022: FDA fully approves Moderna COVID-19 vaccine for individuals ages 18
and older.212
Feb. 18, 2022:
o Following FDA’s apparent decision “not to reconsider authorization” of
Moderna’s COVID-19 vaccine EUA for adolescents ages 12-17, FDA official
Sarah Walinsky sends an invitation for a virtual meeting to Marks and other FDA
officials to discuss how to respond to Moderna’s appeal.213 Walinsky attaches
several items to the meeting invitation including a recent email between FDA
officials discussing how to respond to Moderna and whether FDA should push
Moderna “towards a BLA submission instead of an EUA submission.”214 One
FDA official notes that FDA’s current draft response to Moderna appears to justify
its decision to not authorize the vaccine because “there is not an emergency need
for the Moderna vaccine for this pediatric population, especially given the data
suggesting an increased myocarditis risk compared to the currently available
vaccine for this population (Pfizer).”215
o Walinsky also attaches a preprint version of Szarfman’s forthcoming paper,
“Signaling COVID-19 Vaccine Adverse Events” (which appears to be a finalized
version of the earlier draft titled, “Masking Associated with Early COVID-19
Vaccine Safety Surveillance”), which discusses the increased detection of the
myocarditis signal for the COVID-19 vaccines when using the RGPS method.216
It is unclear why Walinsky circulated the preprint and whether the meeting took
place.
June 17, 2022: FDA issues EUAs for the Moderna COVID-19 vaccines to all children as
young as 6 months old and the Pfizer COVID-19 vaccines to children ages 6 months to 4
years old.217
210 FDA Authorizes Pfizer-BioNTech COVID-19 Vaccine for Emergency Use in Children 5 through 11 Years of Age,
Food and Drug Admin., Oct. 29, 2021, https://web.archive.org/web/20211029200859/https://www.fda.gov/news-
events/press-announcements/fda-authorizes-pfizer-biontech-covid-19-vaccine-emergency-use-children-5-through-
11-years-age.
211 COVID-19 Timeline, Centers for Disease Control and Prevention,
https://www.cdc.gov/museum/timeline/covid19.html.
212 Id.
213 PSI-HHS-000001617192-93; PSI-HHS-000001626344-45; PSI-HHS-000001626448-49.
214 PSI-HHS-000001626344-45.
215 Id.
216 PSI-HHS-000001627327-77.
217 Coronavirus (COVID-19) Update: FDA Authorizes Moderna and Pfizer-BioNTech COVID-19 Vaccines for
Children Down to 6 Months of Age, Food and Drug Admin., June 17, 2022,
28
June 23, 2022: Authors including Szarfman and DuMouchel, publish their paper on
masking titled “Signaling COVID-19 Vaccine Adverse Events” in the journal Drug
Safety.218
July 8, 2022: FDA approves supplemental BLA for the Pfizer COVID-19 vaccine,
authorizing its use for children ages 12 to 15 years old.219
July 12, 2022: Zinderman sends a weekly data mining report to his colleagues at FDA,
including Menschik and Nair, and to a limited number of senior CDC officials, including
Tom Shimabukuro and John Su.220 Based on records reviewed by the Subcommittee, this
appears to be the final weekly data mining report FDA distributes for the COVID-19
vaccines.
July 17, 2022: Szarfman emails her paper entitled “Signaling COVID-19 Vaccine
Adverse Events” to then-FDA Commissioner Robert Califf, noting that it describes
“advances in the data mining (DM) methodology, especially the capacity to unmask
hidden signals due to previously unaccountable confounders[.]”221 Upon receipt of this
and another article, Califf simply replies, “Thanks. These are good.”222
Aug. 5, 2022: In response to an FDA official’s question regarding sending COVID-19
vaccine data mining to CDC, Menschik confirms to his FDA colleagues, “[W]e are no
longer routinely sending COVID data mining to CDC[.]”223 Menschik notes that Nair
had previously discussed this decision with CDC official Shimabukuro.224
Aug. 31, 2022: Forshee writes to Marks and Anderson, notifying them about Szarfman
being a co-author on “Signaling COVID-19 Vaccine Adverse Events.”225 He references
that they are discussing best ways to address “a number of issues with the paper and its
findings[.]”226 Forshee references Szarfman’s earlier efforts in 2021, where she raised
concerns about FDA’s current data mining system.227
https://web.archive.org/web/20220617140937/https://www.fda.gov/news-events/press-announcements/coronavirus-
covid-19-update-fda-authorizes-moderna-and-pfizer-biontech-covid-19-vaccines-
children?utm_medium=email&utm_source=govdelivery.
218 Rave Harpaz et al., Signaling COVID-19 Vaccine Adverse Events, Drug Safety (2022),
https://link.springer.com/article/10.1007/s40264-022-01186-z.
219 COVID-19 Timeline, Centers for Disease Control and Prevention,
https://www.cdc.gov/museum/timeline/covid19.html; Supplement BLA approval, Food and Drug Admin., July 8,
2022, https://www.fda.gov/media/159727/download?attachment.
220 PSI-HHS-000001217046 (with attachment).
221 PSI-HHS-000004461747-48.
222 Id.
223 PSI-HHS-000008266854.
224 Id. After Menschik informs his colleagues that FDA is no longer routinely sending the data mining to CDC, one
FDA official responds, “I was also hoping we would stop doing that at some point!” Id.
225 PSI-HHS-000004594929; Forshee attaches Marks’ September 2021 email to Szarfman’s supervisors Cavazzoni
and Stein. PSI-HHS-000004604858-59.
226 PSI-HHS-000004594929.
227 Id.; PSI-HHS-000004604858-59.
29
Sept. 30, 2022: Nair writes to his FDA colleagues that he suggested to CDC earlier that
summer that “we discontinue the routine regular [weekly data mining] emails.”228 Nair
adds, “[t]his was intended as a timesaving measure and to reduce email traffic.”229 He
informs his colleagues that FDA decided to end its weekly reporting email because “it
had been some time that we had had a data mining alert that required further evaluation,
CDC agreed with this approach.”230 Nair pointed out, however, that “FDA is still
regularly conducting data mining for all approved/authorized vaccines.”231
Nov. 26, 2022: CDC officials discuss how they no longer receive weekly data mining
reports from FDA.232 One official notes, “I think that because of the FOIAs [Freedom
of Information Act requests] we may have asked FDA to stop sending these weekly
data mining outputs.”233 A CDC official eventually writes to Menschik requesting the
most recent data mining report.234
Nov. 28, 2022:
o Menschik emails Nair about how to respond to CDC’s request.235 Menschik notes
that FDA has “concerns about sharing our data mining output externally given
history including over reliance on data mining output[.]”236
o Menschik warns that the data mining results may be “misconstrued” because the
detection of “signals,” or lack thereof, could lead to a false conclusion.237
Because of these issues in addition to the masking limitations, Menschik argues
against sharing the data that CDC requested and states that FDA’s “standard
practice” relies on “assigned reviewers” to evaluate potential safety findings “in
the context of other available data” prior to sharing any safety signal
information.238
Dec. 1, 2022: Nair responds to the CDC official, incorporating many of Menschik’s
points that appear to downplay FDA’s reliance on EB data mining.239 Nair writes,
“Results from data mining are considered hypothesis generating and do not, by
themselves, demonstrate causal associations.”240 He adds that little can be attained from
the presence of a safety signal, or lack thereof, and notes that even the use of the terms
228 PSI-HHS-000001160286.
229 Id.
230 Id.
231 Id.; Per Chairman Johnson’s March 23, 2026 highlighting the FDA’s detection of ischemic stroke following
COVID-19 vaccination, it appears FDA did regularly conduct data mining runs. Letter from Sen. Ron Johnson,
Permanent Subcomm. on Investigations to Robert Kennedy, Secretary, Dep’t of Health and Human Services, March
23, 2026, https://www.ronjohnson.senate.gov/services/files/CA500350-195E-472C-9F26-BE93B290B9D9.
232 PSI-HHS-000002480132.
233 Id. (emphasis added).
234 PSICOVID_00015642-43.
235 Id.
236 Id. (emphasis added).
237 Id.
238 Id.
239 PSICOVID_00014435.
240 Id.
30
“signal and/or safety signal have certain connotations and may trigger actions so we try
not [sic] conflate data mining alerts with signals.”241
Oct. 27, 2023: Nair responds to concerns about masking in FDA’s data mining system
raised by Judy Maro, a Harvard professor working with FDA and CDC officials.242 In his
response, he writes, “We were aware of this limitation before and during the
pandemic. There are many data mining tools and there was some discussion about
utilizing a novel tool to adjust for this. However, we thought it would be problematic
to use a brand new, possibly unvalidated tool in the context of an EUA. We ended up
using the same EBGM [Empirical Bayes Geometric Mean] data mining we use for all
vaccines and has a long history of use rather than take an experimental approach.”243
March 15, 2024:
o Approximately three years after Szarfman raised concerns about the masking
limitation in FDA’s data mining system as a result of the large number of COVID-
19 vaccine reports, FDA officials including Nair, Menschik, and Zinderman
discuss the masking effect.244 Nair writes, “I know in the past we have discussed
one of the possible limitations of data mining currently is the vast number of
VAERS reports from the COVID vaccines may limit our ability to detect
statistical alerts because disproportionality scores may be driven towards the
null.”245 Nair asks his colleagues if they know of any publications discussing the
masking limitation of data mining.246
o In response to Nair, Zinderman writes, “I recall Anna [sic] talking about masking
in the few interactions we had with her, but I don’t remember there being
references.”247
o Following Zinderman’s email, Baer writes to Menschik only and attaches the
“Signaling COVID-19 Vaccine Adverse Events” paper coauthored by Szarfman,
DuMouchel.248 She attaches the paper for Menschik’s consideration because she
believes it could be relevant to Nair’s request for public references to the masking
limitation.249 She follows up later that day providing two specific examples of
articles cited in Szarfman’s and DuMouchel’s paper but ultimately defers to him
on whether to share with Nair as resources.250
o Later, Menschik responds to Nair’s original email and recalls providing the CDC
with language addressing the masking issue for a draft safety article, but “it now
appears that they took it out before publication.”251 This is likely a reference to
the draft language Menschik previously shared with CDC official Su on Sept. 22,
241 Id.
242 PSI-HHS-000001136460-64.
243 PSI-HHS-000001136460 (emphasis added).
244 PSI-HHS-000008261793-94.
245 Id.
246 Id.
247 PSI-HHS-000008261793.
248 PSI-HHS-000008263353.
249 Id.; PSI-HHS-000008261793.
250 PSI-HHS-000008261793.
251 PSI-HHS-000001152913.
31
2021.252 It is unclear whether Menschik provides the articles Baer sent to him.
Sept. 26, 2024:
o Following a discussion about data mining with Oracle contractors, Baer writes to
Menschik about the RPGS algorithm.253 Baer notes that when she compared
results of the RGPS and MGPS models for the Gardasil vaccine, she found that
the RGPS model yielded “significantly higher” numbers than the MGPS model
for certain adverse events.254 She also found that the RGPS numbers were higher
than MGPS numbers for certain adverse events associated with the Pfizer
COVID-19 bivalent vaccine.255
o Baer also mentions Szarfman’s 2022 paper about the masking limitation in data
mining and appears to acknowledge that the lower numbers from the MGPS
model were a result of masking.256 She writes, “I understand the theory behind
masking and trying to adjust for it, but I feel that comprehending the details of the
approach and, importantly, which approach is ‘better,’ is beyond my training and
experience. I think someone with more data mining expertise would have to be
involved in that decision.”257
o In response to Baer, Menschik appears to concur with Baer’s finding and writes
that “in general the [RGPS model] appears way more sensitive in that its scores
are generally higher than corresponding [MGPS model] scores when sampling
different [adverse events].”258 Despite his apparent acknowledgment that RGPS
addresses the masking limitation of MGPS, Menschik appears to express his
preference for continuing to use MGPS.259
252 PSI-HHS-000008268909.
253 PSI-HHS-000008253424.
254 Id. It is unclear what adverse events Baer reviewed that were associated with the Gardasil vaccine.
255 Id. It is also unclear what adverse events Baer reviewed that were associated with the Pfizer COVID-19 bivalent
vaccine.
256 Id.
257 Id.
258 Id.
259 Id.
32
III. Glossary
Key Figures
Anderson, Steven—Director, Office of Biostatistics and Pharmacovigilance, Center for
Biologics Evaluation and Research, FDA.
Baer, Bethany—Division of Pharmacovigilance, Office of Biostatistics and Pharmacovigilance,
Center for Biologics Evaluation and Research, FDA.
Califf, Robert—Commissioner, FDA (Feb. 2022-Jan. 2025).
Cavazzoni, Patrizia—Director, Center for Drug Evaluation and Research, FDA.
DuMouchel, William—Chief Statistician, Oracle.
Forshee, Richard—Associate Director for Research/Acting Deputy Director and later Deputy
Director, Office of Biostatistics and Pharmacovigilance, Center for Biologics Evaluation and
Research, FDA.
Hendrix, Brian—Empirica Signal system, Contractor, Commonwealth Informatics.
Marks, Peter—Director, Center for Biologics Evaluation and Research, FDA.
Maro, Judy—Associate Professor, Department of Population Medicine, Harvard Medical
School and Harvard Pilgrim Health Care Institute.
Menschik, David—Associate Director for Surveillance Informatics, Division of
Pharmacovigilance, Office of Biostatistics and Pharmacovigilance, Center for Biologics
Evaluation and Research, FDA.
Nair, Narayan—Division Director, Division of Pharmacovigilance, Office of Biostatistics and
Pharmacovigilance, Center for Biologics Evaluation and Research, FDA.
Niu, Manette—Acting Branch Chief, Division of Pharmacovigilance, Office of Biostatistics and
Pharmacovigilance, Center for Biologics Evaluation and Research, FDA.
Shimabukuro, Tom—Deputy Director and later Director, Immunization Safety Office, CDC ;
and later, Deputy Director, Influenza Division, Center for Immunization and Respiratory
Diseases, CDC.
Sydnor, James—Empirica Signal system, Contractor, Commonwealth Informatics.
Stein, Peter—Director, Office of New Drugs, Center for Drug Evaluation and Research, FDA.
33
Stockbridge, Norman—Division Director, Division of Cardiology and Nephrology, Center for
Drug Evaluation and Research, FDA.
Su, John—Immunization Safety Office, CDC; and later Deputy Director and Acting Director,
Immunization Safety Office, CDC.
Szarfman, Ana—Medical Officer, Safety Data Mining Developer and Medical Informatics
Analyst, Division of Cardiology, Hematology, Endocrinology, and Nephrology, Center for Drug
Evaluation and Research, FDA.
Walinsky, Sarah—Acting Chief of Staff, Center for Biologics Evaluation and Research, FDA.
Weber, Robert—Product Management Director and Head of Signal Detection and Management
Working Group, Oracle.
Zinderman, Craig—Associate Director for Medical Policy, Office of Biostatistics and
Pharmacovigilance, Center for Biologics Evaluation and Research, FDA.
Key Terms
Empirica Signal—Oracle’s software platform, which utilizes EB data mining, used by FDA for
data mining.260
Empirical Bayesian (“EB”) Data Mining—The data mining method utilized by federal health
agencies to identify statistical associations between products and adverse events.261
Masking—Sometimes referred to as “muting,” masking is a statistical phenomenon in which the
volume of adverse event reports from a similar drug or vaccine product drowns out reports from
other drug or vaccine products, thus distorting the baseline group being compared to the drug or
vaccine of interest being screened.262 This may result in safety signals for adverse events going
undetected.263
Multi-item Gamma Poisson Shrinker (“MGPS”)—The EB data mining algorithm used by
FDA. MGPS was originally developed by William DuMouchel in 1999.264
260 User Guide and Online Help, “About the Oracle Empirica Signal Application,” Oracle, April 1, 2025,
https://docs.oracle.com/en/industries/life-sciences/empirica/2025.4.01/userguide/empirica-signal-application.html.
261 Data Mining at FDA – White Paper, Food and Drug Admin., Aug. 20, 2018, https://www.fda.gov/science-
research/data-mining/data-mining-fda-white-paper.
262 Rave Harpaz et al., Signaling COVID-19 Vaccine Adverse Events, Drug Safety, June 23, 2022,
https://link.springer.com/article/10.1007/s40264-022-01186-z.
263 Id.
264 William DuMouchel, Bayesian Data Mining in Large Frequency Tables, With an Application to the FDA
Spontaneous Reporting System, Am. Stat., (Aug. 1999),
https://www.researchgate.net/publication/254331086_Bayesian_Data_Mining_in_Large_Frequency_Tables_with_an
_Application_to_the_FDA_Spontaneous_Reporting_System_Reply; See also Emeri Potter et al., FDA Adverse
Event Reporting System (FAERS) Essentials: A Guide to Understanding, Applying, and Interpreting Adverse Event
Data Reported to FAERS, Clin. Pharmacol. Ther., May 19, 2025,
https://ascpt.onlinelibrary.wiley.com/doi/10.1002/cpt.3701.
34
Regression-Adjusted Gamma Poisson Shrinker (“RGPS”)—First outlined in a 2012 Oracle
white paper by William DuMouchel, RGPS is an update to MGPS that controls for masking
effects.265
265 William DuMouchel and Rave Harpaz, Regression-Adjusted GPS Algorithm (RGPS), Oracle, Nov. 2012,
https://docs.oracle.com/health-sciences/empirica-signal-811/ESIUG/Regression-Adjusted_GPS_Algorithm.pdf;
Rave Harpaz et al., Signaling COVID-19 Vaccine Adverse Events, Drug Safety, June 23, 2022,
https://link.springer.com/article/10.1007/s40264-022-01186-z.
35
IV. Appendix – Chairman Johnson’s Oversight of COVID-19 Vaccines
Timeline of Chairman Johnson’s Relevant COVID-19 Vaccine Adverse Event Oversight
April 27, 2021: During a meeting with then-National Institutes of Health (“NIH”)
Director Dr. Francis Collins, Sen. Johnson asks him about the alarming number of
adverse event reports in VAERS, particularly the more than 2,900 deaths worldwide
reported within 30-days of vaccination, Collins responded “Senator, people die.”266
June 28, 2021: Letter from Sen. Johnson and Sen. Mike Lee to CDC Director Dr.
Rochelle Walensky and FDA Acting Commissioner Dr. Janet Woodcock raising concerns
about several hundred individuals who reported experiencing an adverse event after
receiving a COVID-19 vaccine.267
July 13, 2021: Letter from Sen. Johnson to Collins, Walensky, and Woodcock regarding
vaccine safety monitoring systems described in the October 2020 VRBAC meeting. Sen
Johnson, writes, “Unfortunately, your agencies’ lack of response to congressional
oversight letters, combined with my discussions with agency officials and individuals
who believe they have experienced vaccine injuries, leads me to believe the
preauthorization safety surveillance hype does not appear to match the agencies’
actual performance.”268
Nov. 2, 2021: Sen. Johnson holds a panel discussion on vaccine mandates featuring
doctors, medical researchers, and vaccine injured.269
Oct. 14, 2021: Letter from Sen. Johnson to Collins, Walensky, Woodcock, Dr. Peter
Marks, Director of the Center for Biologics Evaluation and Research at FDA, and Dr.
Tom Shimabukuro, Deputy Director of the Immunization Safety Office at CDC regarding
federal health agencies’ safety surveillance systems and reports of adverse events
following COVID-19 vaccination.270
Dec. 29, 2021: Letter from Sen. Johnson to Woodcock and Walensky regarding specific
“hot lots” of COVID-19 vaccines associated with increased reports of adverse events.271
Jan. 24, 2022: Sen. Johnson leads a roundtable entitled, “COVID-19: A Second
Opinion.” At the event, Sen. Johnson documents 1.05 million adverse events, and 22,000
deaths connected to the COVID-19 vaccines.272
266 Available at https://www.ronjohnson.senate.gov/services/files/17788FED-A947-4143-8C1B-95C59E60EE87 at
2.
267 Available at https://www.ronjohnson.senate.gov/services/files/EB95C952-175D-405C-8247-AD40FB742DEF.
268 Available at https://www.ronjohnson.senate.gov/services/files/17788FED-A947-4143-8C1B-95C59E60EE87.
269 Available at https://www.ronjohnson.senate.gov/vaccine-side-effects-and-mandates.
270 Available at https://www.ronjohnson.senate.gov/services/files/7EC06E87-9F6F-4E22-8877-8D519CF25A32.
271 Available at https://www.ronjohnson.senate.gov/services/files/F564153D-89FD-40C9-A1B1-8663C22D2F0A.
272 Available at https://www.ronjohnson.senate.gov/vaccine-side-effects-and-mandates.
36
March 1, 2022: Letter from Sen. Johnson to Walensky reiterating outstanding requests
for information on COVID-19 including data on adverse events and vaccine lot variation
data.273
March 23, 2022: Letter from Sen. Johnson to HHS Secretary Xavier Becerra, FDA
Commissioner Dr. Robert Califf, National Institute of Allergy and Infectious Diseases
Director Dr. Anthony Fauci, and Walensky highlighting the over 1 million reports of
adverse events on VAERS and what steps are being taken to address these reports. Sen.
Johnson writes, “As of March 18, 2022, VAERS has received 1,183,495 worldwide
reports of adverse events and 25,641 death reports. Of those deaths, 7,382 (28.8%)
occurred on day 0, 1, or 2 following vaccination.”274
June 23, 2022: Letter from Sen. Johnson to Walensky requesting CDC provide vaccine
safety data, including any EB data mining analyses, created as part of its vaccine safety
monitoring efforts.275
July 25, 2022: Letter from Sen. Johnson to Walensky reiterating request for June 23,
2022 request for vaccine safety data, including EB data mining analyses.276
Sept. 12, 2022: Letter from Sen. Johnson to Walensky in response to CDC producing
only public Proportional Reporting Ratio analyses and failing to respond to requests for
EB data mining analyses.277
Dec. 7, 2022: Sen. Johnson leads a roundtable entitled, “COVID -19 Vaccines: What
They are, How they Work, and Possible Causes of Injuries.” At the event, Sen. Johnson
documents 1.47 million adverse events, and 32,000 deaths connected to the COVID-19
vaccines.278
Jan. 10, 2023: Letter from Sen. Johnson to Walensky reiterating requests for EB data
mining analyses and other vaccine safety data first requested in June 2022.279
April 20, 2023: Letter from Sen. Johnson to Becerra and Carole Johnson, Administrator
of the Health Resources and Services Administration regarding Countermeasures Injury
Compensation Program claims by individuals for COVID-19 vaccine injuries.280
Sept. 5, 2023: Letter from Sen. Johnson to Califf requesting FDA produce all EB data
mining analyses related to the COVID-19 vaccines.281
273 Available at https://www.ronjohnson.senate.gov/services/files/018E3CF9-DCE9-4F0D-836B-DC4DD9866FF2.
274 Available at https://www.ronjohnson.senate.gov/services/files/08942E5E-4220-48A9-B989-9D5947C10BB4.
275 Available at https://www.ronjohnson.senate.gov/services/files/9914278B-A73B-4434-8349-91091138E18B.
276 Available at https://www.ronjohnson.senate.gov/services/files/D48FBED6-BDF3-4FB7-8B24-D52A2EDCE39E.
277 Available at https://www.ronjohnson.senate.gov/services/files/0CBE044E-4F2C-47F2-8272-4DB4F14D3359.
278 Available at https://www.ronjohnson.senate.gov/vaccine-side-effects-and-mandates.
279 Available at https://www.ronjohnson.senate.gov/services/files/AB68101B-CDA4-49F1-8174-4274DDEB0120.
280 Available at https://www.ronjohnson.senate.gov/services/files/A80C5322-F694-4803-BC15-BE3B0D4708D1.
281 Available at https://www.ronjohnson.senate.gov/services/files/0E2CC41F-6202-4124-A7F6-8491E864FB23.
37
Sept. 5, 2023: Letter from Sen. Johnson to Becerra reiterating requests for information
on CICP claims by individuals reporting vaccine injuries following a COVID-19
vaccine.282
Oct. 25, 2023: Letter from Sen. Johnson to Califf, Becerra, CDC Director Dr. Mandy
Cohen, and NIH Acting Director Lawrence Tabak reiterating earlier requests for vaccine
safety information and requesting additional vaccine safety information, including
information on myocarditis and pericarditis after vaccination.283
Dec. 21, 2023: Letter from Sen. Johnson to Becerra, Califf, and Cohen highlighting that
the deaths per million doses of COVID-19 vaccines are approximately 55 times greater
than for deaths per million doses of flu vaccines and requesting all Proportional
Reporting Ratio and EB data mining analyses for the COVID-19 vaccines.284
Jan. 12, 2024: Letter from Sen. Johnson to Califf and Cohen highlighting the issue of
higher rates of adverse events connected to specific manufacture lots of COVID-19
vaccine and reiterating requests from December, including the request for all EB data
mining analysis on COVID-19 vaccines.285
Feb. 26, 2024: Sen. Johnson leads a roundtable entitled, “Federal Health Agencies and
the COVID Cartel: What are They Hiding?”286
Sept. 4, 2024: Letter from Sen. Johnson to Califf and Cohen regarding a study finding
increased all-cause mortality following receipt of the Moderna mRNA COVID-19
vaccine and requesting mortality data related to COVID-19 vaccines from federal health
agencies.287
Nov. 19, 2024: Letter from Sen. Johnson to Becerra, Califf, and Cohen highlighting
heavily redacted records released through the Freedom of Information Act (“FOIA”)
connected to federal health agencies knowledge of the risks of myocarditis and
pericarditis following an mRNA COVID-19 vaccine and requesting unredacted copies of
these records.288
Dec. 5, 2024: Letter from Sen. Johnson to Becerra, Califf, and Cohen highlighting
additional heavily redacted FOIA records connected to federal health agencies’
knowledge of the risks of myocarditis and pericarditis following an mRNA COVID-19
vaccine and requesting unredacted copies of these records.289
282 Available at https://www.ronjohnson.senate.gov/services/files/241E701E-A91E-4B82-B4A8-FF233A3C5C98.
283 Available at https://www.ronjohnson.senate.gov/services/files/8989A37B-49B7-4E87-827F-9BE74D00D06A.
284 Available at https://www.ronjohnson.senate.gov/services/files/0D09CBFB-7A6E-426F-813D-F89DFC4E2EFD.
285 Available at https://www.ronjohnson.senate.gov/services/files/603DE6AC-5F83-4226-8A69-0BB5DCDA6841.
286 Available at https://www.ronjohnson.senate.gov/2024/2/icymi-sen-johnson-leads-roundtable-discussion-federal-
health-agencies-and-the-covid-cartel-what-are-they-hiding.
287 Available at https://www.ronjohnson.senate.gov/services/files/32B052BE-2606-4118-9833-BD0F255867C0.
288 Available at https://www.ronjohnson.senate.gov/services/files/00AAFB3D-72EE-475F-94D5-66708B4AA86D.
289 Available at https://www.ronjohnson.senate.gov/services/files/CCEF4C60-FA50-4A57-804B-D4CA2F835C41.
38
Jan. 28, 2025: Chairman Johnson subpoenas HHS for records connected to COVID-19
vaccines, including EB data mining and vaccine safety monitoring.290
May 21, 2025: PSI releases its interim report titled, “Failure to Warn: How Federal
Health Agencies Downplayed the Risk of Myocarditis and Other Adverse Events
Following COVID-19 Vaccination,” detailing how federal health officials knew about the
risks of myocarditis and pericarditis after an mRNA COVID-19 vaccine and failed to
properly alert the public to those risks.291
Dec. 15, 2025: Letter from Sen. Johnson to Secretary of HHS Robert F. Kennedy, Jr.
regarding Dr. Vinay Prasad’s, the Director of the Center for Biologics Evaluation and
Research, memorandum on deaths of children following a COVID-19 vaccine.292
March 23, 2026: Letter from Sen. Johnson to Secretary of HHS Robert F. Kennedy, Jr.
detailing findings that in late 2022, federal health officials identified, and subsequently
ignored and downplayed, a risk of ischemic stroke in individuals age 65 years and older
who received the Pfizer COVID-19 bivalent booster and raising concerns about masking
in EB data mining used to monitor COVID-19 vaccine safety.293
290 Available at https://www.ronjohnson.senate.gov/services/files/8FAB9531-F799-4067-BA1C-AB8CA182D100.
291 Available at https://www.hsgac.senate.gov/wp-content/uploads/2025.05.21-PSI-Majority-Staff-Interim-Report-
Failure-to-Warn.pdf.
292 Available at https://www.ronjohnson.senate.gov/services/files/AFDAD3A2-D789-46ED-B895-16341762156A.
293 Available at https://www.ronjohnson.senate.gov/2026/3/psi-chairman-johnson-reveals-further-evidence-of-biden-
administration-downplaying-covid-19-vaccine-safety-risk.