Document text
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 1of 27Statistical Analysis Plan (SAP)
for BNT162b2 Vaccine Effecti veness (VE)
Version: 1
Author s:
Srinivas Rao Valluri
Laura Puzniak
MDSCA, Pfizer I nc, New York, NY
Heidi Fischer
Jeff Slezak
Department of Research & Evaluation, Kaiser Permanente Southern California
Date : 9-December -2021
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236158
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 2of 27TABLE OF CONTENTS
LIST OF TABLES ................................ ................................ ................................ ..................... 3
APPENDI CES ................................ ................................ ................................ ........................... 3
1. AMENDMENTS FROM P REVI OUS VERSION(S) ................................ ........................... 4
2. INTRODUCTION ................................ ................................ ................................ ................. 4
3. STUDY DESIGN(S) ................................ ................................ ................................ .............. 5
3.1. KPSC Study C4591014 (CT -24 study ) titled, “Pfizer -BioNTech COVID -19
BNT162b2 Vaccine Effectiveness Study -Kaiser Permanente Southern
California” ................................ ................................ ................................ ................... 5
3.2. Emory Study WI235284 (CT -44 study ) titled, “Respiratory Syncytial Virus
(RSV) in Older Adults and Pregnant Women Study (ROAPS)” ................................ 7
3.3. Bristol Study WI255886 (CT -44 study ) titled, “A Pan- pandemic Acute
Lower Respiratory Tract Disease (LRTD) Surveillance Study (AVONCap)” ........... 8
3.4. Test Negative Design Outcomes and Exposures ................................ ....................... 8
3.5. Full Cohort Study Design ................................ ................................ ........................ 10
3.5.1. Full Cohort Design Outcomes and Exposures (KPSC only ) ...................... 10
4. STUDY OBJECTIVES ................................ ................................ ................................ ........ 12
4.1. Statistical Hy potheses ................................ ................................ ............................. 12
4.1.1. Test Negative Design ................................ ................................ .................. 12
4.1.2. Full Cohort Study Design ( Kaiser only) ................................ ..................... 13
4.2. Statistical decision rules ................................ ................................ .......................... 14
5. ANALYSIS SETS/DAT A SOURCES ................................ ................................ ................ 14
5.1. Full Anal ysis Set ................................ ................................ ................................ .....14
5.1.1. Ful l anal ysis set for KPSC under Test Negative Design ............................ 14
5.1.2. Full anal ysis set for KPSC under Full Cohort Study Design ...................... 15
5.1.3. Full anal ysis set for Emory Study under Test Negative Design................. 15
5.1.4. Full anal ysis set for Bristol Study under Test Negative Design................. 16
5.2. Data Sources ................................ ................................ ................................ ............ 18
5.2.1. KPSC Study ................................ ................................ ................................ 18
5.2.2. Emory Study ................................ ................................ ............................... 18
5.2.3. Bristol Study ................................ ................................ ............................... 18
6. ENDPOINTS AND COV ARIATES ................................ ................................ ................... 18
6.1. Endpoints ................................ ................................ ................................ ................. 18
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236159
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 3of 276.2. Covariates for Test Negative Design and Full Cohort Study Designs .................... 19
7. HANDLING OF MI SSING VALUES ................................ ................................ ................ 20
8. STATI STICAL METHO DOLOGY AND STATI STICAL ANALYSES .......................... 21
8.1. TEST NEGATIVE DESI GN................................ ................................ .................. 21
8.1.1. Descriptive Analy ses................................ ................................ .................. 21
8.1.2. Estimated Crude (Unadjusted) VE ................................ ............................. 21
8.1.3. Estimating Adjusted VE ................................ ................................ ............. 21
8.1.4. Sensitivity Analy ses (Kaiser only ) ................................ ............................. 22
8.1.5. Additional analyses estimating VE for health care workers and other
high risk populations in KPSC study ................................ .............................. 22
8.2. COHORT DESIGN ................................ ................................ ................................ .23
8.2.1. Descriptive Analy ses................................ ................................ .................. 23
8.2.2. Estimated Crude (Unadjusted) VE ................................ ............................. 24
8.2.3. Estimating Adjusted VE ................................ ................................ ............. 24
8.3. Additional Analy tic Elements for Test Negative and Full Cohort Design ............. 24
9. LIST OF TABLES AN D TABLE SHELL S................................ ................................ ........ 24
10. APPENDICES ................................ ................................ ................................ ................... 25
LIST OF TABLES
Table 1. Test-Negative Design ................................ ................................ ............... 12
Table 2. Cohort Design ................................ ................................ ........................... 13
Table 3. Covariates for Test Negative Design and Full Cohort Study Designs .....19
APPENDICES
Appendix 1. Diagnosis and Procedure Codes used in the three protocols ............................... 25
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236160
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 4of 271. AMENDMENTS FROM PREV IOUS VERSION(S)
None.
2.INTRODUCTION
BNT162b2 is a nucleoside -modified m RNA vaccine administered as 2 doses 21 day s apart
that encodes the full- length, membrane -anchored S gly coprotein of SARS -CoV -2 with two
introduced proline mutations to lock it in the prefusion conformation. It was co -developed by
BioNTech SE and Pfizer, I nc. The vaccine showed an acceptable safet y profile in a Phase 1/2
study . In a Phase 3 trial, the vaccine was tolerable and demonstrated 95% efficacy >7 day s
after second dose against COVID -19 in persons without current or prior SARS -CoV -2
infection2.
Pfizer -BioNTech COVID-19 Vacc ine is granted approval b y FDA for active immunization to
prevent coronavirus disease 2019 (COVID -19) caused by severe acute respiratory syndrome
coronavirus 2 (SARS -CoV -2) in individuals 16 y ears of age and older (on 23 Aug 2021) and
has been authorized f or use under an Emergency Use Authorization (EUA) for individuals 5 -
15 and as a booster dose for individuals 18 and over. However, data confirming the
effectiveness of the vaccine outside of the controlled trial setting are still needed. To
evaluate this in a real world setting, Pfizer has undertaken a research collaboration to study
the vaccine effectiveness (VE) of BNT162b2 vaccine against acute respiratory illness
requiring hospitalization due to SARS -CoV -2 infection with the following three external
partners :
1. Kaiser Permanente Southern California (KPSC) [Study C4591014]
2. Emory University (Emory )[Study WI1235284]
3. Bristol University (Bristol) [Study WI255886]
While C4591014 was conducted as a Pfizer sponsored non-interventional study , studies f rom
Emory and Bristol Universities were conducted as non Pfizer sponsored research
collaborations with the respective institutions where the PI is the sponsor .As part of the post
approval commitment under the EUA, Pfizer has a regulatory commitment to eva luate
BNT162b2 vaccine effectiveness using real world data from these 3 protocols. Each of these
three studies have other additional objectives and endpoints outside the scope of the
regulatory commitment ;however , a common objective in these three studies is to evaluate
the VE of BNT162b2 vaccine using a Test Negative Design (TND). Full details of the
primary , secondary , and exploratory objectives and their associated endpoints are described
in greater detail within the respective protocols. For co mpleteness and context, a brief
description of the three protocols along with the COVID -19 related object ives and endpoints
are described below within the respective sections of this Statistical Analy sis Plan (SAP) .
During the review and approval process of the regulatory commitment and the evolving
authorizations and approvals, FDA has provided comments on the above three protocols
necessitating amendments. FDA suggested to submit the amended protocols by the end of
2021 along with the SAPs . Since the pri mary objective is the same in all three protocols,
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236161
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 5of 27evaluating BNT162b2 VE using a TND, Pfizer is preparing and will be submitting a single
SAP that covers the VE analy ses for the above three protocols in response to the regulatory
commitment . Further, theKPSC (C4591014) study will also use a full cohort de sign to
evaluate BNT162b2 VE. Where necessary and appropriate, additional details of the study
design, objectives, endpoints and analy ses undertaken by KPSC using the full cohort design
are also included in this SAP within each section for completeness. Additional analysis plans
covering objectives that are not related to the BNT162b2 commitment may also be developed
and maintained b ythe respective study teams .
Note that this SAP covers only the primary and secondary objectives/endpoints related to the
regulatory commitment on BNT162b2 . Other objectives and endpoints not related to the
regulatory commitment will be described in the project specific SAPs.
Further, adhoc or additional anal yses related t o BNT162b2 as insights emerge from the
rapidly evolving environment willbe documented according to the three respective
institutional policies . Non-COVID endpoints and a nalysesbeyond the ones documented
within this SAP ,if conducted, may be reported either in a separate addendum or within the
final SAP at the end of each study separatel y for transparency .
3.STUDY DESIGN (S)
The study designs for each of the three protocols are described below in brief .For full details
on these three stud y protocols, please refer to the individual protocols C4591014 (for KPSC),
WI235284 (for Emory ), and WI 255886 (for Bristol), respectivel y.
3.1.KPSC Study C4591014 (CT -24 study) titled, “Pfizer -BioNTech COVID-19
BNT162b2 Vaccine Effectiveness Study - Kaiser Permane nte Southern California ”
This is a database anal ysis study of existing healthcare data; no patients will be activel y
enrolled.
This study will be conducted in the Kaiser Permanente Southern California system, a large
integrated healthcare organization with over 4.7 million members who comprise a
socioeconomicall y diverse and broadl y representative population that reflects the
racial/ethnic groups living in Southern California. As of May 14, 2021, KPSC has had over
COVID -19 cases, approximately confirmed patients admitted to the hospital ,
and has vaccinated more than individuals. KPSC has two Regional Laboratories
that process COVID -19 and other specimens. The central reference laboratories receive more
than 50,0 00 specimens per day from the local laboratories and perform over 29 million tests
annually . All laboratories undergo routine qualit y checks to meet or surpass accrediting bod y
specifications.
Vaccine exposure for both study designs (TND and Cohort) considered in KPSC include
initial vaccine series of 2 doses of BNT162b2 received with ≥7 day s between receipt of the
2nd dose and the index date (e.g., admission); partially vaccinated, defined as 1 dose (onl y)
of BNT162b2 received with ≥14 day s between the receipt of the 1st dose and the event date;
ever vaccinated, defined as ≥1 dose of BNT162b2 received with ≥14 day s between the
receipt of the 1st dose and the event date; and >2 doses of BNT162b 2received with ≥14 day s
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
(b) (4)
(b) (4)
(b) (4)
FDA-CBER-2022-5812-0236162
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 6of 27between the receipt of the last dose and the event date. The unexposed group will include
individuals with no record of any COVID -19 vaccination at the time of the event and will
serve as the reference group in the VE anal yses. Additional VE analy ses will compare those
individuals receiving 2 d oses to those receiving >2 doses. All data will be collected from
KPSC electronic health records (EHRs).
Test Negative Design (TND)
A TND study will be used to evaluate the primary objective of this study , i.e., to assess
effectiveness of 2 doses of Pfize r COVID vaccine >7 day s after the second dose against
hospitalization for ARI due to SARS -CoV -2 infection. I t will include all KPSC patients
meeting inclusion criteria admitted to the hospital with ARI after 14 December 2020 (date of
first vaccinations at KPSC), and who had the results of a polymerase chain rea ction ( PCR)
test for SARS- CoV -2. For the secondary objectives estimating VE against emergency
department ( ED)admission, the TND will include KPSC patients meeting inclusion criteria
who present to the ED with (ARI after 14 December 2020 and who receive a PCR test for
SARS -CoV -2. These populations will be used to evaluate additional secondary and
exploratory objectives outlined in the TND proposed objectives, including VE of 1 dose, ≥1
dose, >2 doses and VE for important virus variants.
The index date will be defined as the date of hospitalization or ED admission. Patients can
contribute more than one ARI event to the study if a subsequent ARI event for the same
patient occurred >30 day s after the previous ARI event.
Per KPSC clinical protocol, we expect that ED patients with ARI will be tested for COVID -
19, and in the inpatient setting, all patients with or without ARI will get tested for COVID-19
(to be confirmed with preliminary data) . VE will be estimated separatel y for prevention of
hospitalization (primary outcome) and for prevention of ED presentation without
hospitalization ( a secondary outcome).
Full Cohort Design
A full cohort design will be used to further explore BNT162b2 VE i n the KPSC population.
The cohort stud y will include all KPSC members as of 14 December 2020 (date of first
Pfizer vaccination at KPSC) meeting inclusion criteria for the anal ysis in accordance with
regulatory authorizations. The exposure will be receipt of Pfizer’s COVID -19 vaccine, with
separate relative VEs estimated by number of doses received as stated in the TND above. In
this full cohort design, a patient’s vaccination status, and thus exposure, will change over
time, with all patients entering the cohort as unvaccinated. The outcomes of interest may be
COVID -19 associated hospitalization, ED admission, I CU admission, death, and outpatient
COVID -19 diagnoses (without subsequent hospitalization within 14 day s). As with the
TND, the main outcome of interest will be hospitalization, and the VE of focus will be 2
doses of BNT162b2. The full cohort anal ysis will serve as a secondary analy sis and will
allow for comparability with the TND study . Cohort members will be censored at the time
they disenroll fro m KPSC, die for reasons not related to COVID- 19 (death not within the 30
days following a positive COVID -19 laboratory test), orreceive an y other newly licensed or
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236163
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 7of 27investigational COVID -19 vaccine or prophy lactic agent other than Pfizer’s COVID -19
vaccine .
To control for potential bias and confounding that may exist , individual -level and
neighborhood- level factors will be collected for adjusted analyses. These are factors that
have either been found to be important covariates in previous work, have been id entified in
other risk factor literature, or are variables that may be associated with the exposure as well
as outcome (i.e. prior positive SARS -CoV -2 PCR test, etc.) . Calendar time (days/ weeks/
months depending on data availability ) as a covariate will b e included in the models to adjust
for phase in vaccine rollout, testing practice changes, social distancing impacts, surges, and
potential changes in clinical treatments.
3.2.Emory Study WI235284 (CT-44 study) titled, “Respiratory Syncytial Virus (RSV)
in Older Adults and Pregnant Women Study (ROAPS)”
Emory University , in collaboration with Pfizer, initiated a study in 2018: RSV in Older
Adults and Pregnant Women Study (ROAPS) originall y to evalua te the population- based
incidence of RSV -related hospitalizations in pregnant women and adults ≥50 y ears of age.
When SARS -CoV -2, the cause of COVID -19, emerged as a pathogen in humans in
December 2019 and spread worldwide to become a pandemic over the nex t several months,
ROAPS study team ended study enrollment for Season 2 in mid -March 2020. Earl y data
from the Southern Hemisphere indicated that COVID- 19 and the non -pharmaceutical
interventions that had been implemented (e.g., face masks, social distancin g) as well as
potential changes in the willingness to seek healthcare, have had substantial impact upon the
burden of influenza and the respiratory virus season. I t was unknown whether such
interventions might impact the influenza and RSV seasons in the US . In an effort to continue
to gather data on RSV occurrence in the midst of the pandemic, Emory continued to perform
prospective RSV and viral surveillance (e.g., SARS -CoV -2) among pregnant women, adults
with CHF or COPD, and older adults requiring hospitalization during the pre -planned Season
3 (2020 – 2021); however, modified the approach to rely on data available via medical
records and testing of standard of care specimens, rather than active enrollment, patient
interviews and stud y-specific specimen co llection, as had been done in Seasons 1 and 2. This
change was made to mitigate the challenges that COVID- 19 pandemic -related hospital
precautions posed, including but not limited to, risk of transmission to research staff,
potential shortages of personal protective equipment (PPE), and shortages of respiratory
swabs.
Beginning in the summer of 2020, all admitted patients to EUH and EUHM are receiving
standard of care (SOC) testing for SARS -CoV -2 using a molecular test upon hospital
admission to EUH and EUHM. This provided a unique opportunity to evaluate the VE of
COVID -19 vaccination with BNT162b2 against hospital admission due to ARI in adults and
the protocol was further amended to include VE -related objectives and endpoints (included in
main protoco l as Sub- Study #6). The optimal methodology for VE studies uses the “test -
negative design” (TND) in which vaccination rates among test -positive individuals (“cases”)
are compared with vaccination rates among test -negative individuals (“controls”). A test -
negative case -control study design using a molecular assay is important in estimating VE
accuratel y and rapidly, and can control for differences that might exist due to access to care.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236164
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 8of 27Test-negative controls have been demonstrated to accuratel y estimate VE theoretically and
through retrospective reanaly sis of prospectivel y collected data of vaccine efficacy from
randomized controlled trials. Vaccination status for all cases and controls enrolled in this
study was collected through review of government -issued COVID -19 v accination cards,
medical records from relevant healthcare providers (e.g., primary care, public health
department), health -insurance providers, pharmacies, and any local, state, or national adult
immunization registries. For each potential source of vacci nation, a record of whether
BNT162b2 (or other COVID -19 vaccine) was received, including the date(s) of
administration and the number of doses received, was obtained.
To control for potential bias and confounding that may exist in the absence of randomize d
assignment of vaccine and blinded follow- up and to construct crude OR and VE estimates,
information on time of enrollment, recruitment site, and other potentiall y confounding
sociodemographic, clinical (e.g., comorbidities, history of SARS -CoV -2 infectio n),
behavioral, and lifesty le factors was also collected for use in logistic regression modeling to
assess BNT162b2 VE after adjustment.
3.3.Bristol Study WI255886 (CT -44 study) titled, “A Pan -pandemic Acute Lower
Respiratory Tract Disease (LRTD) Surveillance Study (AVONCap)”
Avon CAP originated as a multi -hospital, prospective surveillance stud y, designed to
determine population -based incidence rates of hospitalized adults ≥18 y ears of age with
community -acquired LRTI (including CAP) in Bristol, England. The i nvolved Bristol
hospitals’ (North Bristol NHS Trust and University Hospitals Bristol) provide near complete
capture of hospital admissions among residents of a well delineated geographic region,
allowing for calculation of population- based incidence rates of LRTI.
As COVID -19 became a pandemic, this study was modified to include objectives to evaluate
the vaccine effectiveness (VE) of BNT162b2. Real world VE estimates for COVID -19
vaccines are needed to demonstrate their effect in general populations as wel l as in risk
groups. These can be achieved using a test negative design (TND) case control anal ysis. For
the purpose of these TND anal yses, cases are defined as individuals testing positive for
COVID -19 up to 14 day s prior to admission or within 3 day s of admission to hospital and
controls are defined as those who had a negative test result in the same timeframe. Almost all
data needed to conduct these anal yses are alread y being collected in this study , including
COVID -19 disease and vaccination status from standard of care records, alongside other
medical history and current illness details. To allow for more complete adjustment for
potential confound ers between the cases and controls, additional information on COVID -19-
related behavioural risk factors are being collected from participants using a standardised
questionnaire, such as occupation, frequency of mask use and social interactions particularly
during periods of lockdown. For the purposes of this SAP, VE evaluation is only for Pfizer’s
BNT162b2 vacci ne.
3.4. Test Negative Design Outcomes and Exposures
The following definitions are emplo yed across all three stud y protocols for the purpose of VE
assessment using this analy tical plan.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236165
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 9of 27Cases: Cases will be defined as those with an y positive laboratory -confir med PCR test from a
sample collected within 14 day s prior to hospital admission through 3 days after a hospital
admission with an yARI symptoms or diagnostic code ( Appendix 1).
Controls: Controls will be defined as those with laboratory confirmed negative COVID-19
(negative COVID -19 test during first 3 day s of hospitalization for primary objective or ED
encounter for secondary objective with an ARI symptoms or diagnostic code ( Appendix 1)
and no positive COVID -19 tests within 14 day s prior to encounter).
Primary Exposure Definition: The exposure of interest is history of vaccination with
BNT162b2. For the primary objective, patients will be considered vaccinated if they have
documented evidence of receiving the second dose of BNT162b2 ≥7 day s before index date
(i.e., defined as the date of hospitalization or ED admission).
Six levels of exposure variable may be assessed:
1.Initial 2- dose vaccination series defined as 2 doses of BNT162b2 received with ≥7 day s
between receipt of the 2nd dose and the index date. This group will serve as the
‘exposed’ group evaluated in the primary objective. Patients who received only 1 dose or
2 doses of BNT162b2 with <7 day s between receipt of the 2nd dose and the index date
will be excluded from analy sis. In sensitivity analy ses, VE will also be calculated for 2
doses of BNT162b2 received with ≥14 day s between receipt of the 2nd dose and the
index date.
2.Partially vaccinated defined as 1 dose (onl y) of BNT162b2 received with ≥14 day s
between receipt of the 1st dose and the index date. This group will serve as the ‘ex posed’
group for the secondary endpoint. Patients who received 2 doses or 1 dose of BNT162b2
with <14 day s between receipt of the 1st dose and the index date will be excluded from
analysis.
3.Ever vaccinated defined as ≥1 dose of BNT162b2 received with ≥14 days between index
date and receipt of the 1st dose. Patients who received 1 dose of BNT162b2 received with
<14 day s between receipt of the 1st dose and the index date will be excluded from
analysis.
4.Greater than 2 doses defined as receiving >2 doses of BN T162b2 with ≥14 day s between
receipt of the most recent dose and the index date.
5.Unvaccinated defined as individuals with no record of any COVID -19 vaccination at
index date . This group will serve as the reference exposure group (i.e., ‘unexposed’
group) i n the VE analy ses.
6. Mixed vaccinated defined as receiving 1 or more doses of BNT162b2 and any other
COVID -19 vaccination or prophy lactic.
The 2 dose exposure group will be considered for the primary objective, while the partially (1
dose), ever vaccinated (≥1 dose) and additional (>2 doses) groups will be considered in
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236166
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 10of 27secondary objectives. Additional comparisons may be made between receiving 2 doses
versus >2 doses or mixed dosing schedules as deemed necessary and will be documented in
the respective protocol specific SAPs.
3.5.Full Cohort Study Design
A full cohort design will be used for secondary objectives to further explore BNT162b2 VE
in the KPSC population.
The cohort stud y will include all KPSC members as of 14 December 2020 (date of first
Pfizer vaccination at KPSC) meeting inclusion criteria for the anal ysis in accordance with
regulatory authorizations. The exposure will be receipt of Pfizer’s C OVID -19 vaccine, with
separate relative VEs estimated by number of doses received. In this full cohort design, a
patient’s vaccination status, and thus exposure, will change over time, with all patients
entering the cohort as unvaccinated. The outcomes of interest may be COVID -19infection,
COVID -19associated hospitalization, ED admission, I CU admission, death, and outpatient
COVID -19 diagnoses (without subsequent hospitalization within 14 day s). As with the
TND, the main outcome of interest will be hos pitalization, and the VE of focus will be 2
doses of BNT162b2. The full cohort anal ysis will serve as a secondary analy sis and will
allow for comparability with the TND study . Cohort members will be censored if they
disenroll from KPSC, die for reasons not related to COVID- 19 (death not within the 30 day s
following a positive COVID -19 laboratory test), receive only another newly licensed or
investigational COVID -19 vaccine or prophy lactic agent other than Pfizer’s COVID -19
vaccine . For additional details o n the cohort design, please refer to the protocol.
3.5.1. Full Cohort Design Outcomes and Exposures (KPSC only)
Main outcome
The main outcome of interest will be COVID -19 hospitalization, which will be defined as a
hospitalization with a positive PCR SARS -CoV -2 test within 14 day s prior and 3 day s after
admission date.
Other outcomes of interest
COVID -19 ICU will be defined as time spent in an intensive care/critical care unit
during a hospital stay with COVID -19 admit as defined above.
COVID -19 ED encounter, which will be defined as an ED encounter with a positive
PCR SARS -CoV -2 test within 14 day s prior and 3 day s after encounter.
COVID -19 death will be defined as death within the 30 days following a positive
COVID -19 laboratory test. Due to lags in compiling cause of death data in our
databases, we will not have cause of death data in time for the study .
Death during hospitalization will also be assessed.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236167
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 11of 27Laboratory -confirmed SARS -CoV -2 infection identified in the outpatient setting,
without a hospitalizatio n in the subsequent 14 days.
Exposures of interest:
The exposure will be receipt of Pfizer’s COVID-19 vaccine, with separate relative VEs
estimated by number of doses received as in the TND design above. In this full cohort
design, a patient’s vaccination status, and thus exposure, will change over time, so VE will
be estimated using time -varying exposures as explained in further detail below.
Partial, Initial 2 dose series, and >2 dose vaccinated : Partial (1 dose) and initial series (2
dose) and greate r than 2 dose (>2 doses) vaccination VE will be estimated using time- varying
exposures, with patients initially entering the cohort as unvaccinated, then contributing
person time to the partial and initial 2 dose series and additional dose(s) vaccinated ex posure
groups as they are receiving the vaccine over time in the real- world setting. Specificall y, a
patient will move to the 1 -dose exposure group once 14 -days have passed following the first
dose, and then to the 2 -dose exposure group once 7 -days have pa ssed following the second
dose and then to the >2 dose exposure group once 14 -days have passed following the third
dose. No requirements on the timing between doses will be applied.
To explore VE durability after 2 and >2 doses, secondary models will further refine the
exposure categories to include time since receipt of dose 2 and >2 doses. As in the main
analysis, patients will still enter the cohort as unvaccinated on 14 December 2020 (date of
first vaccinations at KPSC) and will move from unexposed to the partial and full vaccinated
(2 or 3 or more doses) exposure groups as they are vaccinated over time. Once the second
dose and >2 dose s, is received, we will then code exposure categories as, for example, 30-
day months since reaching vaccination . Each patient will contribute person time to these
groups as the allotted amount of time passes since their second dose. This will allow us to
analyze the relative VE during those different time periods and explore VE durability . If
sample size allows, we will c onduct a similar analy sis looking at only 1 dose, where patients
will be censored from the anal yses when they receive their second dose. To inform our
decisions in choice of cut points for both dose models, we will also estimate changes in
relative VE con tinuously over time by modeling time since vaccination using restricted cubic
splines.
Ever Vaccinated: Relative VE for the Ever Vaccinated ( ≥1 dose) group may be estimated in a
separate anal ysis also using time vary ing exposures. Patients will again enter the cohort as
unvaccinated, then contributing person time to the ever -vaccinated group after receiving dose
1. They will remain in this exposure group regardless of receipt of the 2nd dose or additional
doses of the Pfizer COVID- 19 vaccine.
Unexposed :Individuals with no record of any COVID -19 vaccination or person -time prior
to vaccination among those eventuall y vaccinated.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236168
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 12of 274.STUDY OBJECTIVES
For all three protocols, the primary objective of the study is to estimate vaccine effectiveness
(VE) of 2 dose s of Pfizer’s BNT162b2 vaccine against ARI requiring hospitalization due to
SARS -CoV -2 infection among participants meeting the inclusion criteria of the study . VE
will be evaluated using a TND, including all participants meeting the inclusion criteria of the
study who are admitted to the hospital with ARI after 14 December 2020 (or using date of
first BNT162b2 vaccination at the respective institution), and who receive a test for SARS-
CoV -2. Secondary and exploratory objectives may examine VE for 1 dose va ccination, at
least 1 dose, >2 doses as well as against ED admission, specific variants, mixed dosing
schedules, durability , sequelae, other respiratory pathogens and other populations of interest.
Additionally in KPSC, we will estimate VE using a full cohort design, including all
participants meeting age inclusion criteria of analyses in accordance with the regulatory
authorizations.
4.1.Statistical Hypotheses
4.1.1. Test Negative Design
The full list of endpoints ( i.e.,primary , and secondary ) evaluated using TND are summarized
in the table below.
Table 1. Test- Negative Design
Objective Endpoint /Endpoint Analysis
Primary: (Common to all 3 protocols) Primary:
To estimate the effectiveness of 2 doses of
BNT162b2 against hospitalization for ARI due to
SARS -CoV- 2 infection.
(In Kaiser only ,stratified by age group 16+, 12 -15
and 5 -11 years of age)VEcalculated as 1 minus the odds ratio (OR)
comparing the odds of being vaccinated with 2
doses w ith BNT162b2 for hospitalized cases and
controls, multiplied by 100%.
Secondary: (Common to all 3 protocols) Secondary:
To describe the effectiveness of onl y 1 dose of
BNT162b2 (i.e., partially vaccinated) against
hospitalization for ARI due to SARS -CoV- 2
infection.VEcalculated as 1 minus the OR comparing the odds
of being partially vaccinated with BNT162b2 (only 1
dose) for hospitalized cases and controls, multiplied
by 100%.
To describe the effectiveness of ≥1 dose of
BNT162b2 (i.e., ever vaccinated) against
hospit alization for ARI due to SARS -CoV- 2
infection.VE calculated as 1 minus the OR comparing the odds
of ever being vaccinated ( ≥1 dose) w ith BNT162b2
for hospitalized cases and controls, multiplied by
100%.
To describe the effectiveness of >2 doses of
BNT162 b2 against hospitalization for ARI due to
SARS -CoV -2 infection.VE calculated as 1 minus the OR comparing the odds
of >2 doses w ith BNT162b2 for hospitalized cases
and controls, multiplied by 100%.
To further describe the effectiveness of BNT162b2
against hospitalization stratified by prevalent or
important viral strains BNT162b2 VE estimates stratified by virus variant
(as determined by genotyping or genome
sequencing) and select descriptive analys es described
above by number of doses received
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236169
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 13of 27Table 1. Test- Negative Design
Objective Endpoint /Endpoint Analysis
To evaluate the effectiveness of BNT162b2 against
severe hospitalization -related outcomes (e.g., ICU
admission, mechanical ventilation, and death)BNT162b2 VE estimates against severe outcomes
including ICU a dmission, mechanical ventilation,
and death by number of doses received.
Secondary: (Specific to C4591014) Secondary:
To evaluate overall and variant -specific
effectiveness of BNT162b2 against SARS -CoV -2
infections and COVID- 19 related hospital
admissions by time since vaccination (by month)Monthly VE estimates between variants of interest
using independent Z tests of log hazard ratios.
In particular, for the primary endpoint under H 0, theVE for patients receiving 2 doses of
BNT162b2 are not more than 20% for hospitalized ARI patients (VE<20%). A similar
definition for H 0 is employed for each of the endpoints above where h ypothesis testing is
involved.
4.1.2. Full Cohort Study Design ( Kaiser only)
The full list of endpoin ts (i.e. primary , and secondary ) evaluated using full cohort design in
C4591014 are summarized in the table below.
Table 2.Cohort Design
Objectives Endpoints
Primary: Primary:
1.To estimate the effectiveness of 2 doses of
BNT162b2 against hospitalization due to
SARS -CoV- 2 infection.VE calculated as 1 minus the hazard ratio (HR)
comparing the incidence of 2 doses w ith
BNT162b2 for hospitalization due to SARS -
CoV- 2 infection and not, multiplied by 100%.
Secondary: Secondar y:
1. To estimate the effectiveness of 2 doses of
BNT162b2 against ED admission (w ithout
subsequent hospitalization) ED admission
due to SARS -CoV -2 infection.VE calculated as 1 minus the HR comparing the
incidence of 2 doses w ith BNT162b2 for ED
admission due to SARS -CoV -2 infection and not,
multiplied by 100%.
2.To estimate the effectiveness of 2 doses of
BNT162b2 against ICU admission due to
SARS -CoV- 2 infectionVEcalculated as 1 minus the HR comparing the
incidence of 2 doses w ith BNT162b2 for ICU
admission due to SARS -CoV -2 infection and not,
multiplied by 100%.
3.To estimate the effectiveness of 2 doses of
BNT162b2 against death due to SARS -
CoV- 2 infectionVE cal culated as 1 minus the HR comparing the
incidence of (2 doses w ith BNT162b2 for death due
to SARS -CoV -2 infection and not, multiplied by
100%.
4.To estimate the effectiveness of 2 doses of
BNT162b2 against COVID -19 outpatient
visits (without subsequent hos pitalization
within 14 days) due to SARS -CoV- 2
infectionVE calculated as 1 minus the HR comparing the
incidence of 2 doses w ith BNT162b2 for COVID -19
outpatient visits (without subsequent hospitalization
within 14 days) due to SARS -CoV- 2 infection and
not, multiplied by 100%.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236170
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 14of 27Table 2.Cohort Design
Objectives Endpoints
5.To describe the effectiveness of only 1 dose
of BNT162b2 (i.e., partially vaccinated)
against hospitalization, ED admission, ICU
admission, death, and outpatient visits
(without subsequent hospitalization within
14 days) due to SARS -CoV -2 infection.VE calculated as 1 minus the HR comparing the
incidence of only 1 dose of BNT162b2 (i.e., partially
vaccinated) for hospitalization, ED visit, death, and
COVID -19 outpatient visits (without subsequent
hospitalization within 14 days) due to SARS -CoV -2
infection and not, multiplied by 100%.
6.To describe the effectiveness of ≥1 dose of
BNT162b2 (i.e., ever vaccinated) against
hospitalization, ICU admission, ED
admission, death, and outpatient visits
(without subsequent hospitalization within
14 days) due to SARS -CoV -2 infection.VEcalculated as 1 minus the HR comparing the
incidence ≥1 dose of BNT162b2 (i.e., ever
vaccinated) for hospitalization, ED visit, death, and
COVID -19 outpatient visits (without subsequent
hospitalization within 14 days) due to SARS -CoV -2
infection and not, multiplied by 100%.
7.To describe the effectiveness of >2 doses of
BNT162b2 against hospitalization, ED
admission, ICU admission, death, and
outpatient visits (without subsequent
hospitalization within 14 days) due to
SARS -CoV -2 infection.VE calculated as 1 minus the HR comparing the
incidence of >2 doses of BNT162b2 for
hospitalization, ED visit, death, and COVID -19
outpatient visits (without subsequent hospitalization
within 14 days) due to SARS -CoV- 2 infection,
multiplied by 100%.
In particular, for t he primary endpoint under H 0, the VE for patients receiving 2 doses of
BNT162b2 are not more than 20% for hospitalized ARI patients (VE<20%). A similar
definition for H0 is employ ed for each of the endpoints above where hy pothesis testing is
involved.
4.2.Statistical decision rules
No adjustments for multiple comparisons will be made.
5.ANALYSIS SETS/ DATA SOURCES
5.1.Full Analysis Set
The full anal ysis sets (FAS) for each of the protocols is described below.
5.1.1. Full analysis set for KPSC under Test Negative Design
Inclusion criteria
1.KPSC patients who are admitted to the hospital (primary and some secondary objective s)
with ARI (ARI ; ICD codes listed in Appendix 1) after 14 December 2020 (date of first
vaccinations at KPSC), and who receive a PCR test for SARS- CoV -2.
2.For the secondary objectives estimating VE against ED admission, the TND will include
KPSC patients who present to the ED with ARI after 14 December 2020, and who
receive a PCR test for SARS- CoV -2.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236171
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 15of 273.We will include membership requirement of 1 year prior to index date, which is defined
as the date of hospitalization or ED admission (allowing a 31- day administrative gap), to
facilitate accurate capture of comorbid conditions.
Exclusion criteria
Patients who receive only another newl y licensed or investigational SARS -CoV -2 vaccine or
COVID -19 proph ylactic agent other than Pfizer’s COVID -19 vacci ne prior to hospitalization
(or ED) will be excluded from the anal ysis. When estimating VE for BNT162b2 vaccination,
patients receiving another newl y licensed or investigational SARS -CoV -2 vaccine or
COVID -19 proph ylactic agent other than Pfizer’s COVID -19vaccine prior to hospitalization
or ED will be excluded from the anal ysis. Patients will also be excluded if the index date is
within certain time windows from vaccination date, outlined further in the exposure
section 3.4above .
Age restrictions for anal yses will be made based on the objective and aligned with regulatory
authorization/approvals; i.e., separate VE anal yses will be conducted for persons aged 12-15
and 5 -11years .
5.1.2. Full analysis set for KPSC under Full Cohort Study Design
Inclusion criteria
1. All KPSC members as of 14 December 2020 (date of first Pfizer vaccination at
KPSC).
2.For the cohort study , patients must have at least 1 year of membership (allowing a
31-day administrative gap) prior to 14 Decemb er 2020 (index date, date
vaccinations first began at KPSC) to facilitate accurate capture of comorbid
conditions.
Exclusion criteria
There will be no exclusion criteria for the cohort design, however patients will be censored
for receiving an y other newl y licensed or investigational SARS -CoV -2 vaccine or COVID -
19 prophy lactic agent other than Pfizer’s COVID- 19 vaccine.
5.1.3. Full analysis set for Emory Study under Test Negative Design
Inclusion Criteria
1.Age 18 years or older.
2.Admitted to hospital for ARI *at a participating site.
3.Previously provided a standard of care specimen (NP or nasal swab) on this hospital
admission or willing and able to provide specimen (NP or nasal swab) and comply
with all data collection requested.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236172
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 16of 274.Capab le of providing informed consent (or LAR capable and willing to give informed
consent), which includes compliance with the requirements and restrictions listed in
the protocols. In the case a LAR is not available, a waiver of consent is requested.
*ARI for study enrollment will be defined as:
a.ARI symptoms (nasal congestion, rhinorrhea, sore throat, hoarseness, new or
increased- from -baseline cough, sputum production, dyspnea, wheezing) OR
b. Admitting diagnosis suggestive of ARI (Pneumonia, Upper respir atory infection,
Bronchitis, Influenza, Cough, Asthma, Viral respiratory illness, Respiratory
distress, AND/OR Respiratory failure.
The ARI definition for analysis will include the WHO definition (e.g., acute symptoms of
fever and cough) and variations sin ce about 15% of patients admitted with COVID -19 do
not have fever or cough.
Exclusion criteria
1.Previous enrollment in this study within the past 30 day s.**
2.Any contraindication to have a NP or nasal swab (if specimen was not collected as
SOC).
** Thus, pa tients can contribute >1 ARI event to the study if a subsequent ARI event for the
same patient occurred >30 days after the previous event.
5.1.4. Full analysis set for Bristol Study under Test Negative Design
Screening Inclusion criteria
Patients must meet all th e following inclusion criteria to be eligible for enrolment:
1.Aged ≥18 y ears of age
2.Patients with illness with following 2 characteristics:
a.Acute illness (i.e., present for 28 day s or less); AND
b.Evidence of acute LRTD:
i.Patients with current or suspected C OVID -19 or previous proven COVID -19
within last 28 day s OR
ii.Clinical or radiologic diagnosis of pneumonia or an acute LRTI OR
iii.New onset or worsening of ≥2 of following 8 LRTD sy mptoms or clinical
findings:
1. fever (>38.0°C) or h ypothermia (<35.5°C ) before or within 24 hours of
enrolment;
2. pleuritic chest pain;
3. cough (including nocturnal only );
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236173
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 17of 274. sputum production or purulence;
5. dyspnea (shortness of breath) including orthopnea or on exertion only ;
6. tachypnea (respiratory rate ≥20/min) do cumented by healthcare
professional;
7. abnormal auscultatory findings suggestive of LRTD (e.g.,
crepitations/rales or evidence of pulmonary consolidation including
dullness on percussion, bronchial breath sounds, wheezing, or
egophon y);
8. radiologic fin ding that is consistent with L RTD, including pneumonia,
and/or acute congestive heart failure (e.g., pleural effusion, increased
pulmonary density due to infection, the presence of alveolar infiltrates
(multilobar, lobar or segmental) containing air bronch ograms, or
interstitial oedema).
Screening Exclusion criteria
Patients meeting an y of the following criteria will not be included in the study:
1.Any patient who develops signs and s ymptoms of LRTD after being hospitalized for ≥48
hours (either at current ho spital, another transferring hospital, or a combination of these),
unless admitted with current, previous proven, or suspected COVID -19 infection.
2.Previously enrolled participants readmitted ≤7 day s after discharge for their study
qualify ing admission, unl ess admitted with current, previous proven, or suspected
COVID -19 infection
3.At the time of enrolment, an L RTD -related diagnosis has been excluded or another
diagnosis confirmed (for example, patient was found to have fever and tachy pnoea due to
an intraabdominal process such as cholecy stitis)
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236174
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 18of 275.2.Data Sources
5.2.1. KPSC Study
All data will be collected from KPSC electronic health records. This is a database analy sis
study of existing healthcare data; no patients will be activel y enrolled. We will collect data
including vaccination status and dates of vaccination, COVID- 19 testing and outcomes,
comorbidities, prior healthcare utilization, other vaccinations, demographic data, and other
data from the EHR.
5.2.2. Emory Study
Data from the ROAPS study are collected from a variet y of sources including patient
interview (behavioral/social characteristics, patient -reported vaccine history), electronic
medical records (standard of care testing results, clinical data, medical history data, etc.),
Emor y laboratory data (including whole genome sequencing results for COVID- positive NP
swabs if/when available), and the Georgia Registry of Immunization Transactions and
Services (GRITS) (documented vaccine registry history ).
5.2.3. Bristol Study
Adults with L RTD will be screened using population -level surveillance at study hospitals,
and collection of SOC data will be performed on all LRTD events, including from SOC
laboratory tests. Documented or suspected COVID- 19 will fulfil the study eligibility criterion
for LRTD, thus if patients meet the other stud y entry criteria (such as age etc), L RTD
patients will be offered participation in the enhanced diagnostic testing portion of this study
with informed consent, which will involve collection of urine and respiratory samples. These
samples will be used for study -specific testing and, if necessary , for COVID -19,
pneumococcus, and RSV tests if not available from SOC records for an y reason. A short
patient questionnaire on COVID -related risk behaviours will also be admin istered.
Information about the additional pneumococcal, SARS- CoV -2 and RSV infection testing will
be integrated with the population- level surveillance data to allow for more accurate
population -based estimates of vaccine-preventable pneumococcal and COVID- 19 and RSV -
related LRTD incidence. The epidemiologic data generated from the stud y will serve as the
baseline for future vaccine effectiveness studies, including for current and possibly future
SARS -CoV -2 vaccines.
6.ENDPOINTS AND COVARI ATES
6.1.Endpoints
Test Negative Design
All endpoints considered in this SAP are related to VEof the BNT162b2 vaccine onl y and
are detailed in section 4.1.1 for the TND . The definitions for cases and controls used to
assess VE for BNT162b2 vaccine are also described in section 3.4 along with the definitions
for exposure. The 2 dose exposure group will be considered for the primary objective, while
the partially (1 dose), ever vaccinate d (≥1 dose) and greater than 2 doses (>2 doses) groups
will be considered in the secondary and exploratory objectives.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236175
FDA-CBER-2022-5812-0236176
FDA-CBER-2022-5812-0236177
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 21of 278. STATISTICAL METHODOL OGY AND STATISTICAL ANALYSES
Statistical Analy sesand methods adopted are described below.
8.1.TEST NEGATIVE DESIGN
The anal yses below will be done separatel y based on the objective, number of doses and
endpoints.
8.1.1. Descriptive Analyses
Proportion of hospitalized patients with ARI where SARS -CoV -2 was identified, as well as
the proportion of patients who received 0, 1, 2, and >2 doses of BNT162b2 will be
descriptivel y summ arized by case and control status. A verage and median time between the
receipt of the first and second dose of BN T162b2 among patients who received 2 doses,
between the receipt of the second and third doses of BNT162b2 among patients who received
3 doses, and between December 14, 2020 (beginning of vaccination) and receipt of the first
and last dose of BNT162b2 will be descriptivel y summarized. Additionally , age, gender,
race/ethnicity , clinical characteristics, and severit y (ICU admission, mechanical ventilation ,
death), and other characteristics as collected and described in the respective protocols of an y
patients who received BNT162b2 and tested positive for SARS -CoV -2will be summarized .
8.1.2. Estimated Crude (Unadjusted) VE
Odds ratios and corresponding 95% confidence intervals (CIs) of BNT162b2 vaccination
(>2 doses, 2 doses, 1 dose, ≥1 dose) for cases and test -negative controls will be estimated
using the G eneralized Estimating Equation (G EE) with logit link function or a logistic
regression model as appropriate. VE will be calculated as 1 −OR multiplied by 100%.
8.1.3. Estimating Adjusted VE
In addition to constructing crude OR and VE estimates, logistic regression model or GEE
with logit link function will be performed to assess BNT162b2 VE after adjustment for the
potentially confounding factors will be performed. Potential confounders to include in
models will be selected based on prior knowledge and empirical findings through the model
building process. Findings from the phase- 3 studies, expert opinion, and published studies of
clinical/biologic factors will be used to generate a list of candidate variables. These will be
assessed for their availability and ability for adjusting crude results, including an examination
of their distributions and missingness. Bivariate associations of potential confounders with
the outcome, exposure, and each other will be examined. For variables with suggestions of
imbalance with respect to exposure or outcome, the association between exposure and
outcome will be stratified by categories of the potential confounder to look for differences
across strata (potential effect modification) and influence on summary estimates of
association (confounding). The variables will ultimately be selected for inclusion in final
adjusted models based on a combination of a priori decisions and a qualitative assessment of
the empirical relationships. Sensitivity analy ses will assess the robustness of cut -point
selections and groupings and consider tightly confounded variables and their col linearity .
The results of the multivariable model should be consistent with the stratified anal yses.
Corresponding 95% CIs will be calculated using the Wald method. A GEE estimator will be
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236178
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 22of 27used with a robust sandwich variance estimator to account for clus tering introduced by
variables measured at the neighborhood level. In addition to the fully adjusted model that
includes relevant covariates, univariate VE results will be presented for each independent
variable that is assessed for potential confounding . A 2-sided alpha of 0.05 will be used in all
analyses.
8.1.4. Sensitivity Analyses (Kaiser only)
1.It is possible that patients that present with a COVID -19-like illness or a COVID -19
diagnosis are not tested for SARS- CoV -2 within 3 day s of hospital admission but ar e
rather tested later in their hospital stay . If this were the case, we would want to
expand the requirement of a COVID -19 diagnostic test bey ond 3 day s following
admission. To investigate the possibility of late testing, we will present data on the
distri bution of COVI D-19 tests at time since admission for those admitted for
respiratory infections. If a meaningful number of patients are tested >3 days after
hospital admission, a sensitivity analy sis to examine VE without time restrictions on
testing follow ing admission may also be included.
2.KPSC will u se and develop N atural Language Processing (NLP) algorithms to
estimate actual date of s ymptom onset of COVID- 19 sy mptoms. Sy mptom onset will
then be considered to define exposure status at the time of a qualify ing event (or to
censor a patient if they experience an event before 14- days after the first dose of 7 -
days after the second).
8.1.5. Additional analyses estimating VE for health care workers and other high risk
populations in KPSC study
We know that the vaccine roll- out is following a tiered strategy , for example, with healthcare
workers with direct patient contact being vaccinated first. The logic supporting tiered vaccine
eligibility is based on COVID -19 risk, with highest risk populations prioritized first. To
account for differing risk profiles of vaccinated individuals over time, we will account for
calendar time in our model. To explore whether controlling for calendar time (length of time
to be determined b y sample size) is sufficient to address possible biases in VE , we will
assess three options:
1.Flag healthcare worker, or other sub -population status (gold standard).
a.This will require complete and reliable identification of healthcare worker or
other sub- population status, for example, L TCF-resident, in the EHR. This is
the preferred approach.
b.Analy ses will then be stratified so that both cases and controls will come from
the same sub- population.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236179
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 23of 272.Assess and compare VE of models stratified b y time periods.
a.KPSC documentation of the dates of transitions between tiers will be used to
create categories of vaccine distribution by time (Hea lthcare worker s/LTCF
residents only , 65+, etc.),
b.Using these categories, we will perform stratified analy ses for each phase in
the vaccine rollout and examine any differences in relative VE between time
strata.
3.If, through our anal yses in part 2 or as the result of additional clinical input, we
determine our inability to identify patients eligible for vaccination during certain
vaccination tiers will result in unobserved confounding that will materiall y affect the
reliability of our VE estimates, we will limit VE anal yses to certain time periods of
interest for which we know vaccinations were restricted to a particular tier –in
particular we may drop analy ses focused on the time period only healthcare workers
were vaccinated and focus on time periods were vaccination is more widespread .
The Bristol and Emory studies have a questionnaire included in the study whereby patient
risk factors are collected and assessed as covariates.
8.2.COHORT DESIGN
As in the TND, we will include those with and without prior COVID -19 diagnose s.
8.2.1. Descriptive Analyses
Proportion of patients who receive d 0, 1, 2, and >2 doses of BNT162b2 will be descriptivel y
summarized. Average and median time between the receipt of the first and second dose of
BNT162b2 among patients who received 2 doses, between the receipt of the second and third
doses of BNT162b2 among patients who received 3 doses, and between December 14, 2020
(beginning of vaccination) and receipt of the first and last dose of BNT162b2 will be
descriptivel y summarized. Overall inciden ce of the outcomes of interest will be calculated b y
dividing the number of outcome cases by the total number of person -years. We will also
provide incidence estimates by age, gender, race/ethnicity , clinical characteristics, and other
covariates described in the covariate table in section 6.2. A dditionally ,changes in the
clinical and demographic composition of the vaccinated and non -vaccinated population over
time may be explored , as this will also change with vacc ine phase integration. Characteristics
of those who test positive for COVID -19 and those without COVI D-19 will be presented.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236180
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 24of 278.2.2. Estimated Crude (Unadjusted) VE
Vaccine effectiveness ( VE)will be estimated as (1 –Hazard Ratio)*100%, and the hazard
ratio (H R) will be estimated using a Cox proportional hazard model with c orresponding 95%
CIs calculated. Vaccination status will be time -varying as described previously in
section 3.5.1 .
8.2.3. Estimating Adjusted VE
Adjusted haza rd ratios (HRs) and 95% CIs will be estimated b y including age, sex, race, and
other covariates listed in section 6.2 in Cox proportional hazard regression models. Potential
confounders to include in models will be selected based on prior knowledge and empirical
findings through the model building process. Findings from the phase -3 studies, expert
opinion, and published studies of clinical/biologic factors will be used to generate a list of
candidate variables. The se will be assessed for their availability and ability for adjusting
crude results, including an examination of their distributions and missingness. Bivariate
associations of potential confounders with the outcome, exposure, and each other will be
examine d. For those with suggestions of imbalance with either exposure or outcome, the
association between exposure and outcome will be stratified by categories of the potential
confounder to look for differences across strata (potential effect modification) and influence
on summary estimates of association (confounding). The variables will ultimately be
selected based on a combination of the a priori decisions and a qualitative assessment of the
empirical relationships. Sensitivity analy ses will assess the rob ustness of cut -point selections
and groupings and consider tightly confounded variables and their collinearity. The results
of the multivariable model should corroborate the knowledge gained from the stratified
analyses.We will control for calendar week in all models. Robust variance will be computed
to account for clustering introduced b y neighborhood level variables. Vaccine effectiveness
(VE) will be estimated as (1 –adjusted HR)*100%, and the hazard ratio (HR) will be
estimated using Cox proportional hazard model with corresponding 95% CI s calculated.
8.3.Additional Analytic Elements for Test Negative and Full Cohort Design
1.Provide descriptive statistics and determine VE stratified by virus variants determined
to be important or prevalent based on sequencing anal yses.
2.Determine VE of BNT162b2 stratified by various patient characteristics (e.g., age,
sex, chronic medical conditions, receipt of influenza vaccine). Table 2has the full
list of proposed stratified anal yses
9. LIST OF TABLES AND T ABLE SHELLS
Please see separate Excel document which outlines proposed table shells for these analy ses.
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236181
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 25of 2710. APPENDICES
Appendix 1. Diagnosis and Procedure Codes used in the three protocols
Acute Respiratory Infection Diagnosis codes
ICD-10 code ICD-10 definition
J12.0 Adenoviral pneumonia
J12.1 Respiratory syncytial virus pneumonia
J12.2 Parainfluenza virus pneumonia
J12.81 Pneumonia due to SARS -associated coronavirus
J12.82 Pneumonia due to coronavirus disease 2019
J12.3 Human metapneumovirus pneumonia
J12.89 Other viral pneumonia
J12.9 Viral pneumonia, unspecified
J13 Pneumonia due to Streptococcus pneumoniae
J18.1 Lobar pneumonia, unspecified organism
J15.0 Pneumonia due to Klebsiella pneumoniae
J15.1 Pneumonia due to Pseudomonas
J14 Pneumonia due to Hemophilus influenzae
J15.4 Pneumonia due to other streptococci
J15.4 Pneumonia due to other streptococci
J15.3 Pneumonia due to streptococcus, group B
J15.4 Pneumonia due to other streptococci
J15.20 Pneumonia due to staphylococcus, unspecified
J15.211 Pneumonia due to Methicillin susceptible Staphylococcus aureus
J15.212 Pneumonia due to Methicillin resistant Staphylococcus aureus
J15.29 Pneumonia due to other staphylococcus
J15.8 Pneumonia due to other specified bacteria
J15.5 Pneumonia due to Escherichia coli
J15.6 Pneumonia due to other Gram -negative bacteria
A48.1 Legionnaires' disease
J15.8 Pneumonia due to other specified bacteria
J15.9 Unspecified bacterial pneumonia
J15.7 Pneumonia due to Mycoplasma pneumoniae
J16.0 Chlamydial pneumonia
J16.8 Pneumonia due to other specified infectious organisms
B25.0 Cytomegaloviral pneumonitis
A37.01 Whooping cough due to Bordetella pertussis with pneumonia
A37.11 Whooping cough due to Bordetella parapertussis with pneumonia
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236182
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 26of 27ICD-10 code ICD-10 definition
A37.81 Whooping cough due to other Bordetella species with pneumonia
A37.91 Whooping cough, unspecified species with pneumonia
A22.1 Pulmonary anthrax
B44.0 Invasive pulmonary aspergillosis
J17 Pneumonia in diseases classified elsewhere
B77.81 Ascariasis pneumonia
J17 Pneumonia in diseases classified elsewhere
J18.0 Bronchopneumonia, unspecified organism
J18.8 Other pneumonia, unspecified organism
J18.9 Pneumonia, unspecified organism
J10.00 Influenza due to other identified influenza virus with unspecified type of
pneumonia
J10.01 Influenza due to other identified influenza virus with the same other identified
influenza virus pneumonia
J10.08 Influenza due to other identified influenza virus with other specified pneumonia
J11.00 Influenza due to unidentified influenza virus with unspecified type of pneumonia
J11.08 Influenza due to unidentified influenza viru s with specified pneumonia
J12.9 Viral pneumonia, unspecified
J10.1 Influenza due to other identified influenza virus with other respiratory
manifestations
J11.1 Influenza due to unidentified influenza virus with other respiratory
manifestations
J10.2 Influenza due to other identified influenza virus with gastrointestinal
manifestations
J10.81 Influenza due to other identified influenza virus with encephalopathy
J10.82 Influenza due to other identified influenza virus with myocarditis
J10.83 Influenz a due to other identified influenza virus with otitis media
J10.89 Influenza due to other identified influenza virus with other manifestations
J11.2 Influenza due to unidentified influenza virus with gastrointestinal manifestations
J11.81 Influenza due to unidentified influenza virus with encephalopathy
J11.82 Influenza due to unidentified influenza virus with myocarditis
J11.83 Influenza due to unidentified influenza virus with otitis media
J11.89 Influenza due to unidentified influenza virus with ot her manifestations
J09.X1 Influenza due to identified novel influenza A virus with pneumonia
J09.X2 Influenza due to identified novel influenza A virus with other respiratory
manifestations
J09.X3 Influenza due to identified novel influenza A virus with gastrointestinal
manifestations
J09.X9 Influenza due to identified novel influenza A virus with other manifestations
J10.08 Influenza due to other identified influenza virus with other specified pneumonia
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236183
Protocols C4591014, WI255886, and WI235284 Statistical Analysis Plan
PFIZER CONFIDENTIAL
CT24- WI-GL03 -RF03 2.0 Non-Interventional Statistical Analysis Plan For Secondary Data Collection Study
01-Jun-2020
Page 27of 27ICD-10 code ICD-10 definition
J10.1 Influenza due to other identified influen za virus with other respiratory
manifestations
J09.X9 Influenza due to identified novel influenza A virus with other manifestations
J09.X1 Influenza due to identified novel influenza A virus with pneumonia
J09.X2 Influenza due to identified novel influenza A virus with other respiratory
manifestations
J09.X3 Influenza due to identified novel influenza A virus with gastrointestinal
manifestations
J09.X9 Influenza due to identified novel influenza A virus with other manifestations
090177e198d5f266\Approved\Approved On: 10-Dec-2021 16:40 (GMT)
FDA-CBER-2022-5812-0236184