Document text
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 1
NON -INTERVENTIONAL (NI) STUDY STATISTICAL ANALYSIS PLAN (SAP)
Study Information
Title Pfizer -BioNTech COVID -19 Vaccine
Exposure during Pregnancy: A Non -
Interventional Post -Approval Safety Study
of Pregnancy and Infant Outcomes in the
Organization of Teratology Information
Specialists (OTIS)/MotherToBaby
Pregnancy Registry
Protocol number C4591022
Statistical Analysis Plan version identifier 12.0
Date 30 November 2021 05 May 2022
EU Post Authorization Study (PAS)
register number EUPAS42869
Active substance COVID -19 mRNA Vaccine is single -
stranded, 5’ -capped messenger RNA
(mRNA) produced using a cell-free in vitro
transcription from the corresponding DNA
templates, encoding the viral spike (S)
protein of SARS -CoV -2
Medicinal product Pfizer -BioNTech COVID -19 Vaccine
(BNT162b2)
Research question and objectives Research Question: Is the risk of pregnancy
and infant safety outcomes increased among
pregnant women in the Organization of
Teratology Information Specialists
(OTIS)/MotherToBaby Pregnancy Registry
who were vaccinated with the Pfizer -
BioNTech COVID -19 vaccine during
pregnancy compared wit h those who did not
receive any COVID -19 vaccine during
pregnancy ?
Study Objective
• To assess whether pregnant women
who received the Pfizer -BioNTech
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236332
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 2 COVID -19 vaccine during pregnancy
experienced increased risk of
pregnancy and infant safety
outcomes, inclu ding major
congenital malformations,
spontaneous abortion, elective
termination/abortion, stillbirth,
preterm delivery, small for
gestational age, and small for age
postnatal growth at one year of age,
relative to pregnant women who
received no COVID -19 vaccines
during pregnancy.
Authors Christina Chambers, PhD, MPH
Professor of Pediatrics
School of Medicine
University of California San Diego
9500 Gilman Drive, MC 0828
La Jolla, CA 92093
Tel: +1 858 -246-1704
Email: [email protected]
Ronghui (Lily) Xu , PhD
Professor of Math ematics
University of California San Diego
9500 Gilman Drive, MC 0828
La Jolla, CA 92093
Email: [email protected]
University of California S an Diego
9500 Gilman Drive, MC 0828
La Jolla, CA 92093
Email @health.ucsd.edu
Gordon Honerkamp Smith, MS
Statistician
University of California San Diego
9500 Gilman Drive, MC 0828
La Jolla, CA 92093
Email: [email protected]
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
(b) (6)
(b) (6)
FDA-CBER-2022-5812-0236333
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 3 1. TABLE OF CONTENTS
1. TABLE OF CONTENTS ................................ ................................ ................................ .......3
2. LIST OF ABBREVIATIONS ................................ ................................ ................................ 5
3. AMENDMENTS AND UPDATES ................................ ................................ ....................... 7
4. RATIONALE AND BACKGROUND (SUMMARY) ................................ ......................... 8
5. RESEARCH QUESTION AND OBJECTIVE ................................ ................................ ......8
6. STUDY DESIGN (SUMMARY) ................................ ................................ .......................... 9
7. STUDY POPULATION (SUMMARY) ................................ ................................ .............. 10
7.1. Inclusion Criteria ................................ ................................ ................................ .....10
7.2. Exclusion Criteria ................................ ................................ ................................ ....10
7.3. Follow -Up ................................ ................................ ................................ ............... 10
8. STUDY SIZE ................................ ................................ ................................ ....................... 11
8.1. Sample Size ................................ ................................ ................................ ............. 11
8.2. Power Calculations ................................ ................................ ................................ ..11
9. DATA SOURCE (SUMMARY) ................................ ................................ ......................... 13
9.1. Maternal Interviews ................................ ................................ ................................ .13
9.2. Medical Records and General Pediatric Evaluation ................................ ................ 15
10. VARIABLES ................................ ................................ ................................ ..................... 16
10.1. Timing Variables ................................ ................................ ................................ ...16
10.2. Identification of Exposure and Comparator ................................ ......................... 16
10.3. Classification of Pregnancy and Infant Outcomes ................................ ................ 20
10.3.1. Major Congenital Malformations ................................ ............................. 20
10.3.2. Spontaneous abortion ................................ ................................ ................ 22
10.3.3. Elective termination/abortion ................................ ................................ ...22
10.3.4. Stillbirth ................................ ................................ ................................ ....22
10.3.5. Preterm delivery ................................ ................................ ........................ 22
10.3.6. Small for gestational age at birth ................................ .............................. 22
10.3.7. Small for age postnatal growth at one year of age ................................ ....22
10.4. Demographic and Clinical Characteristics ................................ ............................ 23
11. MISSING DATA ................................ ................................ ................................ ............... 26
12. STATISTICAL METHODS AND DATA ANALYSIS ................................ ................... 27
12.1. Data Cleaning and Preparation of Datasets ................................ ........................... 27
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236334
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 4 12.2. Demographic and Baseline Characteristics ................................ ........................... 27
12.3. Primary Analyses ................................ ................................ ................................ ..27
12.3.1. Major Congenital Malformation ................................ ............................... 30
12.3.2. Spontaneous Abortion, Stillbirth, and Preterm Delivery .......................... 31
12.3.3. Elective Termination/Abortion ................................ ................................ .33
12.3.4. Small for Gestational Age at Birth and Small for Age Postnatal
Growth at One Year of Age ................................ ................................ ............. 33
12.4. Secondary Analyses ................................ ................................ .............................. 34
12.4.1. Stratified/Subgroup Analyses ................................ ................................ ...34
12.4.2. Individual Dose Effects ................................ ................................ ............ 34
12.4.3. Evaluation for a Pattern of Major Congenital Malformatio ns .................. 38
12.4.4. Lost to Follow -Up................................ ................................ ..................... 38
12.5. Sensitivity Analyses ................................ ................................ .............................. 39
12.6. Interim Analyses ................................ ................................ ................................ ...39
12.7. Analysis Software ................................ ................................ ................................ .40
13. REFERENCES ................................ ................................ ................................ .................. 40
14. LIST OF TABLES ................................ ................................ ................................ ............. 45
15. LIST OF FIGURES ................................ ................................ ................................ ........... 45
ANNEX 1. LIST OF STAND -ALONE DOCUMENTS ................................ ......................... 45
ANNEX 2. ADDITIONAL INFORMATION ................................ ................................ ......... 45
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236335
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 5 2. LIST OF ABBREVIATIONS
Abbreviation Definition
BMI body mass index
CDC Centers for Disease Control and Prevention
CI confidence interval
cm centimeter
COVID -19 Coronavirus disease 2019
EUA Emergency Use Authorization
FDA Food and Drug Administration
HCP health care provider
HR hazard ratio
IPTW inverse probability of treatment weighting
kg kilogram
LMP last menstrual period
m meter
MACDP Metropolitan Atlanta Congenital Defects Program
MI multiple imputation
MICE multivariate imputation by chained equations
mRNA messenger ribonucleic acid
MSM marginal structural model
NCHS National Center for Health Statistics
OR odds ratio
OTIS Organization of Teratology Information Specialists
PASS Post-Authorization Safety Study
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236336
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 6 Abbreviation Definition
PCR polymerase chain reaction
PDA patent ductus arteriosus
PFO patent foramen ovale
PMC postmarketing commitment
RNA ribonucleic acid
RR risk ratio
SAP Statistical Analysis Plan
SARS -CoV -2 severe acute respiratory syndrome coronavirus 2
SMD standardized mean difference
SOP standard operating procedure
Tdap tetanus, diphtheria, and acellular pertussis
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236337
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 7 3. AMENDMENTS AND UPDATES
None.
Amendment
Number Date SAP
Section(s)
Changed Summary of Amendment(s) Reason
1 05 May 2022 6.1
Inclusion
Criteria Removed "age 18 years or older" from
the inclusion criteria To allow for enrollment of
individuals who are <18 years of
age as per CBER request given
vaccine authorization/approval in
younger ages.
9.1 Timing
Variables Updated the definition of trimester s to
≤13, 13.1 -≤26, 26
To align with the definitions used
for the study
11 3.1
Major
Congenital
Malformati
ons Updated formula for inverse probability
weighting . Updated number of
bootstrap samples from 10,000 to 200 To fix an error and prevent
negative values ; to align with the
current method of using multiple
imputation
Throughout
document Fixed minor typos and references. To fix errors
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236338
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 8 NOTE: In this document, any text taken directly from the Non -Interventional (NI) study
protocol is italicized.
4. RATIONALE AND BACKGROUND (SUMMARY)
Pfizer and BioNTech have partnered to develop a novel messenger ribonucleic acid (mRNA)
vaccine (Candidate BNT162b2) against a novel coronavirus (severe acute respiratory
syndrome coronavirus 2 [SARS -CoV -2]) for the prevention of associated coronavirus dis ease
2019 (COVID -19). The United States Food and Drug Administration ( FDA) initially granted
Emergency Use Authorization (EUA) for the Pfizer -BioNTech COVID -19 vaccine two -dose
primary series on 11 December 2020 in individuals 16 years of age and older, and approved
the vaccine for this population on 23 August 2021 ( FDA, 2021a ; FDA, 2021b ). FDA
expanded the EUA on 10 May 2021 to include children 12 -15 years of age, and on 29
October 2021 to include lower -dose vaccine administration for children 5 -11 years of age
(FDA, 2021c ; FDA, 2021d ). The EUA was further amended on 12 August 2021 to include the
administration of a third primary series dose in certain immunocompromised individuals 12
years of age and older, and on 22 September 2021 to allow for use of a single booster dose at
least six months after completion of the primary series in certain populations ( FDA, 2021e ).
Available data suggest that pregnant women who become infected with COVID -19 may be
more likely to be hospitalized and may be at increased risk of preterm delivery ( MMWR,
2020 ). Pfizer is conducting a Phase 2/3 clinical trial of the safety and immunogenicity of the
Pfizer -BioNTech COVID -19 vaccine in pregnant women. While the current product labeling
communicates that data are insufficient, the Pfizer -BioNTech COVID -19 vaccine may be
received by pregnant women when they and their healthcare providers believe that
risk/benefit considerations favor its use. As of 25 October 2021 , more than 169,000 women
reported to the Centers for Disease Control and Prevention’s (CDC) V -safe surveillance
system that they were vaccinated during pregnancy (CDC_2021a ). Therefore, information
regarding the real -world safety of vaccination during pregnancy is essential from a public
health perspective.
This observational study described in the corresponding study protocol is being conducted to
evaluate pregnancy and infant safety outcomes among pregnant women enrolled in an
established North American pregnancy registry who were exposed to the Pfizer -BioNTech
COVID -19 vaccine. This non -interventional study is designated as a Post -Authorization
Safety Study (PASS) and is a postmarketing commitment (PMC) to the FDA.
This statistical analysis pla n (SAP) provides a comprehensive and detailed description of
statistical approaches and techniques to analyze the data for the study.
5. RESEARCH QUESTION AND OBJECTIVE
Research Question: Is the risk of pregnancy and infant safety outcomes increased among
pregnant women in the Organization of Teratology Information Specialists
(OTIS)/MotherToBaby Pregnancy Registry (OTIS Pregnancy Registry) who were vaccinated
with the Pfizer -BioNTech COVID -19 vaccine during pregnancy compared with those who
did not receive an y COVID -19 vaccine during pregnancy?
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236339
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 9 Study Objective
• To assess whether pregnant women who received the Pfizer -BioNTech COVID -19
vaccine during pregnancy experienced increased risk of pregnancy and infant safety
outcomes, including major congenital malformations, spontaneous abortion, elective
termination/abortion, stillbirth, preterm delivery, small for gestational age, and small
for age postnatal growth at one year of age, relative to pregnant women who received
no COVID -19 vaccines during pregnancy.
6. STUDY DESIGN (SUMMARY)
This proposed study is a prospective, observational cohort study of pregnancy and infant
safety outcomes in pregnant women in the OTIS Pregnancy Registry who received the
Pfizer -BioNTech COVID -19 vaccine any time from one month before the first day of the last
menstrual period (LMP) to the end of pregnancy. The comparator cohort includes pregnant
women who received no COVID -19 vaccines within one month before the first day of LMP
to end of preg nancy. The pregnancy outcomes are major congenital malformations,
spontaneous abortion, elective termination/abortion, stillbirth, preterm delivery, and small
for gestational age. The infant outcome is small for age postnatal growth at one year of age.
The target sample size for the study is 2000 pregnant women : 1100 pregnant women in the
Pfizer -BioNTech COVID -19 vaccine exposure cohort and 900 pregnant women in the
COVID -19 vaccine -unexposed comparator cohort . The main measures of effect are
unadjusted and adjusted risk ratios (RRs) and 95% confidence intervals (CIs) comparing the
Pfizer -BioNTech COVID -19 vaccine exposed cohort to the comparator cohort for the
outcomes of major congenital malformations, small for gestational age, and postnatal
growth; and u nadjusted and adjusted hazard ratios (HRs) and 95% CIs comparing the
Pfizer -BioNTech COVID -19 vaccine exposed cohort to the comparator cohort for the
outcomes of spontaneous abortion, elective termination/abortion, stillbirth, and preterm
delivery . Seconda ry analyses will be conducted to address the potential bias in prenatal
diagnostic testing procedures to detect major congenital malformations performed prior to
versus after enrollment, the separate effects of trimester of vaccine exposure for endpoints
other than major congenital malformations, individual dose effects, potential bias in those
who are lost to follow -up, and the effect of prior COVID -19 infection. Sensit ivity analyses
will be conducted to evaluate major congenital malformations identified in all pregnancies
including those not ending in at least one live birth , but excluding those lost to follow -up; in
pregnancies with any abnormal ultrasound findings pri or to enrollment; in pregnancies
restrict ed to those only those enrolled in the first trimester; and in an expand ed comparator
group includ ing those vaccinated with the Pfizer -BioNTech COVID -19 vaccine only in the
second and/or third trimester. A sensitivi ty analysis for the endpoint of preterm delivery will
be conducted stratified by those delivering following spontaneous labor versus labor
induction or delivery by cesarian section.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236340
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 10 7. STUDY POPULATION ( SUMMARY )
7.1. Inclusion Criteria
Individuals must meet all of the following inclusion criteria to be eligible for inclusion in the
study:
• Residence in the US or Canada
• Age 18 years or older
• Enrolled in the OTIS Pregnancy Registry and currently pregnant on or after 11
December 2020 (i.e., date of EUA for the Pfizer -BioNTech COVID -19 vaccine)
7.2. Exclusion Criteria
Individuals meeting any of the following criteria will not be included in the study:
• Previous entry into this study for a prior pregnancy
• Receipt of any vaccine other than the Pfizer -BioNTech COVID -19 vaccine, influenza
vaccine, or Tdap vaccine (e.g., Moderna or Johnson & Johnson COVID -19 vaccines,
human papillomavirus vaccine, hepatitis B vaccine, etc.) from one month before the
first day of LMP up to and including end of pregnancy
• Exposure to known human teratogens dur ing pregnancy within one month before the
first day of LMP up to and including end of pregnancy
• Known pregnancy outcome at time of study enrollment (e.g., positive prenatal
diagnostic test results for a major congenital malformation prior to study entry)
7.3. Follow -Up
The study entry date, i.e., the follow -up start date, for pregnant women in the OTIS
Pregnancy Registry is the study enrollment date. Information is collected on their pregnancy
to date and they are then followed for the duration of their pregnancy. In addition, infants
will be followed for potential safety events through their first year of life .
Follow -up will end at the earliest of t he following events:
• Lost to follow -up, i.e., an enrolled individual who withdraws or who fails to complete
the outcome interview despite a standard number of telephone attempts and attempt
to contact by mail as per study procedure manual within one year o f the participant’s
estimated due date
• Receipt of any vaccine other than the Pfizer -BioNTech COVID -19 vaccine, influenza
vaccine, or Tdap vaccine (e.g., Moderna or Johnson & Johnson COVID -19 vaccines,
human papillomavirus vaccine, hepatitis B vaccine, etc. ) during pregnancy
• Occurrence of spontaneous abortion, elective termination/abortion, or stillbirth
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236341
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 11 • End of follow -up (i.e., one -year post -partum)
• End of study period
• Infant death
• Maternal death during pregnancy
8. STUDY SIZE
8.1. Sample Size
For the target sample size of 1800 participants enrolled in the study , recruitment goals are
set at 1100 participants in the Pfizer -BioNTech COVID -19 vaccine exposure cohort, and 900
participants in the comparator cohort. The target sample size for the Pfi zer-BioNTech
COVID -19 vaccine exposure cohort was increased from an initial sample size of 900 to 1100
to allow for increased capture of women receiving a booster dose during pregnancy. The
rationale for the target sample sizes in each cohort was based on several considerations for
feasibility of enrollment. These included early trends in referrals of COVID -19 vaccinated
pregnant women to the OTIS Pregnancy Registry, combined with a reasonable time period
for recruitment of pregnant women and the necessary time to collect outcome data to one
year postpartum. An additional consideration was reasonable statistical power to detect
differences for each outcome of interest. Balance in the cohort numbers by trimester of
exposure in the vaccine -exposed cohort and b y trimester of enrollment in both cohorts will be
monitored on a monthly basis, and overall balance addressed by adjusting recruitment
activities as needed. It is not possible to predict if the recruitment rates will be equal in all
years, and therefore, s ample size is based on estimates that may require revision as the study
progresses.
As women will be eligible to enroll at any time in pregnancy, the gestational weeks at
enrollment is expected to vary from 2 weeks to 41 weeks. Based on the gestational weeks at
enrollment, only the portion of the overall sample enrolled prior to 20 weeks’ gestation will
be eligible for the analysis of spontaneous abortion. We estimate based on prior experience
that half the overall sample will enroll prior to 20 weeks. Simi larly, only the subset of the
sample enrolled prior to 37 weeks’ gestation will be eligible for the analysis of preterm birth.
It is estimated that 95% of the overall sample will enroll before 37 weeks. For the outcome of
major congenital malformations , the main subset eligible for this assessment will be
restricted in the exposed cohort to women who received at least one dose of the Pfizer -
BioNTech COVID -19 vaccine from one month prior to the first day of LMP to the end of the
first trimester. Pregnant women who received the Pfizer -BioNTech COVID -19 vaccine only
during their second or third trimester will be excluded from the analysis. Based on previous
experience in OTIS studies, the estimated lost to follow -up rate is 5% ( Chambers et al, 2016 ).
8.2. Power Calculations
Based on these assumptions, for the outcome of major congenital anomalies , it is estimated
that 85% will result in live birth after exclusion of pregnancy losses (10%) in women
enrolled in the first half of pregnancy, and lost -to-follow -up (5%). For spontaneous abortion,
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236342
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 12 it is estimated that half of the overall sample will enroll prior to 20 weeks’ gestation. For
preterm delivery, it is estimated that 80% of the overall sample will enroll prior to 37 week s’
gestation and will end in a singleton live birth. For the outcome of small for gestational age,
it is estimated that 85% of the overall sample will end in a singleton live birth.
Baseline birth prevalence, incidence rates, and incidence proportions of m ajor congenital
malformations, preterm delivery, small for gestational age, and small for age postnatal
growth at one year of age, respectively, are based on previous OTIS Pregnancy Registry
studies and on general population data. Table 1 gives the power f or various detectable
relative risks (RRs)/hazard ratios (HRs) for two -sided alpha level of 0.05 in the comparisons
of the exposed cohort to the comparator cohort for the range of background risks of the
outcomes of interest.
Table 1. Sample Size and Power for a Specified Effect Size
Outcome N in Exposed
Cohort N in Comparator
Cohort Birth Prevalence/
Incidence in
Comparator Cohort Detectable
Relative Risk/
Hazard Ratio Power1
Major congenital
malformations2 311 765 3%3 2.1 65.8%
2.5 86.7%
2.7 92.6%
Spontaneous
abortion 550 450 10%4 1.5 66.6%
1.7 90.1%
1.8 95.6%
Preterm delivery 880 720 10%5 1.4 69.1%
1.5 85.6%
1.6 94.7%
Small for
gestational age 935 765 10%6 1.4 71.7%
1.5 87.6%
1.6 95.8%
Small for age
postnatal growth
at one year of age 935 765 10%6 1.4 71.7%
1.5 87.6%
1.6 95.8%
1. Arcsine transformation using pwr.2p.test() in R package pwr’ to obtain effect size h, where ℎ=2∗
𝑠𝑖𝑛−1√𝑝1−2∗𝑠𝑖𝑛−1√𝑝2 and p1 = event rate in the exposed cohort, p 2 = event rate in the
comparison cohort, assuming 2 sided alpha = 0.05.
2. Among livebir ths.
3. CDC, 2017.
4. Avalos et al., 2012.
5. Ferre et al., 2016.
6. CDC, 2017; Nellhaus, 1968; Olsen et al., 2010.
NOTES
i. Major Congenital Malformations
• N in exposed cohort for Major congenital malformations based on assumptions that 366 exposed
to at least one dose from 1 month prior to LMP through the first trimester, and of these 10% will
be lost to spontaneous abortion or stillbirth, and 5% lost to fol low-up; 85% of the 366 enrolled
will thus yield 311 eligible for the analysis.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236343
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 13 Table 1. Sample Size and Power for a Specified Effect Size
Outcome N in Exposed
Cohort N in Comparator
Cohort Birth Prevalence/
Incidence in
Comparator Cohort Detectable
Relative Risk/
Hazard Ratio Power1
• N in comparator cohort = 900 enrolled x 85% resulting in at least one live birth = 765.
ii. Spontaneous Abortion
• N in exposed cohort for spontaneous abortion based on estimate that ½ will be exposed and enroll
in the first 20 weeks of gestation and thus be at risk of spontaneous abortion until 20 completed
weeks; N of 1100 enrolled and exposed x 50% enrolled and exposed prior to 20 weeks = 550.
• N in comparator cohort = 900 of which 50% will be enrolled prior to 20 weeks = 450.
iii. Preterm Delivery
• N in exposed cohort based on 95% of the 1100 enrolled and exposed to at least one dose prior to
37 weeks, 10% pregnancy loss, and 5% lost to follow -up for an overall estimated 80% eligibl e for
the analysis; 1100 enrolled x 80% = 880.
• N in comparator cohort based on same assumptions; 900 enrolled x 80% = 720.
iv. Small for Gestational Age
• N in exposed cohort based on 1100 enrolled and exposed to at least one dose from 1 month prior
to LMP up th rough the end of pregnancy, 10% pregnancy loss, and 5% lost -to-follow -up; 1100 x
85% = 935.
• N in comparator cohort based on same assumptions; 900 x 85% = 765.
v. Small for Age Postnatal Growth at One Year of Age
• N in exposed cohort based on 1100 enrolled and exposed to at least one dose from 1 month prior
to LMP up through the end of pregnancy; 10% pregnancy loss and 5% lost -to-follow -up; 1100 x
85% =935.
• N in comparator cohort based on same assumptions; 900 x 85% = 765.
9. DATA SOURCE (SUMMARY)
As part of the OTIS Pregnancy Registry protocol, data are collected using maternal
interview(s) and medical record review. To supplement the interim and pregnancy outcome
interviews and to improve recall, participants are given a pregnancy exposure diary to
record any additional exposures (medications, vaccinations, vitamins, etc.) or events as the
pregnancy progresses.
Data are recorded on hard copies of forms and these records are retaine d by OTIS at the
OTIS Research Center. These forms are considered the primary data sources for studies and
can be adapted to add new data elements. Data from these forms are extracted and entered
into a customized OTIS study database.
9.1. Maternal Interviews
The maternal interviews are conducted by telephone at the intake/enrollment interview, then
1-2 interim interviews (depending on gestational age of the pregnant woman at enrollment) ,
and lastly the pregnancy outcome interview. Data collected from each of th ese interviews
include the following:
• Intake/Enrollment Interview
o Pregnancy history, including major congenital malformations, genetic
disorders, number of live births, and multiple gestations
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236344
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 14 o Current health history
o Pre-pregnancy weight and height
o Socioeconomic and demographic information including maternal and
paternal occupation, education and ethnicity
o Income category
o Any COVID -19 vaccine exposure prior to and during pregnancy , including
dates, scheduled dose, and manufacturer
o Vaccine use from one month prior to the first day of LMP and throughout
pregnancy
o Current medication use, both prescription and over the counter
o Other environmental or occupational exposures
o Alcohol, tobacco, caffeine and illicit drug use
o Current pregnancy complicatio ns including illnesses
o Family history of adverse pregnancy outcomes, including major congenital
malformations and genetic disorders
o Names and addresses of health care providers
o COVID -19 symptoms, treatments, and testing results
o Referral source
• Interim Inte rviews I and II at 20 -22 and 32 -34 weeks’ gestation (if enrolled at those
times)
o Update of data since last interview, including records of pregnancy exposures
(medications, vaccinations, vitamins, supplements, and other prescription and
over-the-counter pr oducts ), results of prenatal tests, events of interest ( e.g.,
pregnancy complications, illnesses, pregnancy end prior to the expected due
date), and contact info rmation
o COVID -19 symptoms, treatments, and testing results
• Pregnancy Outcome Interview at 0 to 6 weeks after the expected due date (or at an
interim interview point or earliest convenient time for the participant if pregnancy
has ended)
o For women with live born infants:
▪ Date of delivery, hospital location and mode of delivery
▪ Sex, birth weight, leng th and head circumference
▪ Apgar scores
▪ Description of delivery or birth complications including
malformations
▪ Type and length of hospital stay for pregnant women and their infants
▪ Delivering physician’s and infant physician’s names and addresses
▪ Method of infant feeding
▪ Pregnancy weight gain
▪ COVID -19 symptoms, treatments, and testing results
▪ Additional exposures and results of prenatal tests occurring since the
previous interview
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236345
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 15 o For women with spontaneous or elective abortions:
▪ Date and type of outcome
▪ Hospital location if applicable
▪ Prenatal diagnosis
▪ Pathology results if available
▪ COVID -19 symptoms, treatments, and testing results
▪ Additional exposures and results of prenatal tests occurring since the
previous interview
o For women with stillborn infants:
▪ All of the above for women with spontaneous or elective abortions
▪ Sex
▪ Delivery or birth complications including malformations
▪ Birth size
▪ Autopsy results if available
9.2. Medical Records and General Pediatric Evaluation
Medical records for pregnant wom en and their infants are captured at birth and again for
the infant at one year of age to supplement information self -reported by the participant
related to vaccine exposure, outcomes, prenatal tests, and medical history.
Medical records from the prenatal care provider, the hospital of delivery, any specialty
provider, and the pediatric care provider will be requested and abstracted for exposure,
outcome and covariate/confounder data. A standard pediatric questionnaire will be completed
by the physician res ponsible for the care of each live born infant at or near one year of age.
Data collected from medical records and the pediatric questionnaire include:
• Exposure to Pfizer -BioNTech COVID -19 vaccine including dose number and dates;
• Pregnancy outcome;
• Prenatal tests and results;
• Pregnancy complications;
• Mode of delivery;
• Birth weight, length and head circumference of infant;
• Apgar scores;
• Length and type of hospital stay;
• Major congenital malformations identified in the fetus or infant up throu gh one year
of age;
• Postnatal growth measures for the infant up to one year of age (measurements
between 9 to 15 months of age are considered valid; if multiple measurements are
available, the evaluation of weight, length and head circumference that is closest to
one year of age will be used ;
• COVID -19 infection, testing and treatments.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236346
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 16 To supplement maternal report of vaccination, a copy of the COVID -19 vaccine record is
also requested from participants who have been vaccinated.
10. VARIABLES
Variables for the exposures, outcomes, demographics, and clinical characteristics of interest
are included below. Data on these variables will be collected via maternal interview and
medical record review per standard process described in the OTIS Pregnancy Registry
protoco l. Detailed operational definitions are provided below.
10.1. Timing Variables
• The estimated date for the first day of LMP is based on maternal report or medical
record . If the first day of LMP and the menstrual cycle length are known, the first day
of LMP is us ed to calculate the estimated date of confinement or due date. However,
if a first trimester ultrasound has been performed, and the estimated due date by that
ultrasound differ s by seven days or more from the first day of LMP estimate, the first -
trimester ultrasound -derived date will be used. If there is no first trimester ultrasound,
but an ultrasound performed in the second trimester has an estimated due date that
diffe rs by14 days or more, or a third trimester ultrasound estimated due date that
differs by 21 days or more from the estimate by first day of LMP , the ultrasound -
derived date will be used. The earliest available ultrasound in pregnancy is used to
determine if any adjustment in due date calculated by first day o f LMP is necessary.
When the first day of LMP and/or cycle length are unknown or when a prenatal
ultrasound estimates a gestational week that is discrepant according to obstetric
guidelines, the due date calculated by the earliest available ultrasound is used
(ACOG, 2017 ).
• Weeks’ g estation is defined as the number of weeks from the estimated first day of
LMP which is counted as day 0 and calculated as:
(Current date – estimated date of the first day of LMP )/7.
• The definition of trimesters is as follows:
o First trimester: 30 days prior to the first day of LMP to <≤13 weeks’ gestation
o Second trimester: 13 .1 weeks’ gestation to <≤26 weeks’ gestation
o Third trimester: ≥26 weeks’ gestation .
• The start date for exposure ascer tainment during pregnancy is defined as one month
prior to the first day of LMP and calculated as the estimated date of the first day of
LMP minus 30 days.
10.2. Identification of Exposure and Comparator
Vaccine exposure data are obtained by maternal report and/or medical record to classify
Pfizer -BioNTech COVID -19 vaccine exposure status. The two sources of information are
used in order to minimize misclassification of exposure status. Detailed informatio n
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236347
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 17 regarding vaccine exposure status of participants is obtained through the maternal
interviews including at enrollment, at interim timepoints during pregnancy, and at pregnancy
outcome interview (0 to 6 weeks after the expected due date or end of pregnanc y).
Participants are directly queried about the specific vaccines they received, including
information on the gestational timing, dates of exposure, and manufacturer . Vaccine
exposures are coded using the Slone Drug Dictionary. Maternal report that a COVID -19
vaccine was received is complemented by requesting a copy of the COVID -19 vaccine
record. In addition, medical records from the obstetric provider, hospital of delivery, and
any specialty provider are reviewed (when available) for all participants.
A participant is classified as unexposed if she reports that she did not receive any COVID -19
vaccine and her medical record (when available) shows no indication that she received any
COVID -19 vaccine. A participant is classified as exposed if she reports in the maternal
interviews that she received the Pfizer -BioNTech COVID -19 vaccine or if there is supporting
documentation in the medical record for receipt of the Pfizer -BioNTech COVID -19 vaccine.
If maternal report indicates no receipt of a COVID -19 vaccine but the medical record
indicates discordance such that the Pfizer -BioNTech COVID -19 vaccine was administered,
the medical record would supersede maternal report and the participant would be classified
as Pfizer -BioNTech COVID -19 vaccine -exposed. Moreover, where maternal report indicates
that the Pfizer -BioNTech COVID -19 vaccine was received but documentation in the medical
record is discordant (i.e., no indication in medical record that COVID -19 vaccine was
administered), the maternal report would supersede medical record for classification as
exposed.
Table 2 below provides a description of variables for exposures to be included in this study.
Each pregnancy and infant outcome will be analyzed separately, and the following exposure
and comparator cohorts wi ll be defined separately for each outcome analysis ( depending on
a participant’s timing of vaccination/study enrollment as it relates to the at -risk period for an
outcome ):
• Pfizer -BioNTech COVID -19 Vaccine (Exposure)
– Receipt of at least one dose of the Pfi zer-BioNTech COVID -19 vaccine during
the study period at any time from one month before the first day of LMP up to
and including end of pregnancy
• COVID -19 Vaccine -Unexposed (Comparator)
– Receipt of no COVID -19 vaccines during the study period within one mo nth
before the first day of LMP up to and including end of pregnancy
Table 2. Variables Related to Defining the Exposure, Comparator, Outcomes,
Exclusions, and Date/Time
Variable Role Data source(s) Operational definition
Exposure to the
Pfizer -
BioNTech Exposure Maternal report Maternal report , vaccine record documentation,
or medical record documentation of exposure to at
least one dose of the Pfizer -BioNTech COVID -19
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236348
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 18 Table 2. Variables Related to Defining the Exposure, Comparator, Outcomes,
Exclusions, and Date/Time
Variable Role Data source(s) Operational definition
COVID -19
vaccine Vaccine record (e.g.,
COVID -19 vaccine
or yellow card)
Medical record vaccine any time from one month prior to first day
of LMP to end of pregnancy .
No COVID -19
vaccine
exposure during
pregnancy Comparator Maternal report
Medical record Maternal report of no exposure to, as well as n o
medical record documentation of , any COVID -19
vaccines at any time from one month prior to first
day of LMP to end of pregnancy
Weeks’
gestation Time Maternal report
Vaccine record (e.g ,
COVID -19 vaccine
or yellow card)
Medical record Number of weeks (rounded to the nearest 0.1
decimal ) from the first day of LMP to date of
interest calculated as:
(Date of interest – estimated date of
the first day of LMP ) / 7
The date of interest is based on maternal report ,
vaccine record, or medical record depending on
which variable the date is for and which source
was used to assign a value to that variable.
If the date of interest (e.g., vaccine exposure date)
is in the 30 days prior to the first day of LMP, the
gestational week would be between -0.1 and -4.0.
Weeks’
gestation at time
of receipt of the
Pfizer -
BioNTech
COVID -19
vaccine Timing of
exposure Maternal report
Vaccine record (e.g.,
COVID -19 vaccine
or yellow card)
Medical record Weeks’ g estation (using calculation in row above)
for date of ‘Exposure to the Pfizer -BioNTech
COVID -19 vaccine’ based on the vaccine dose
administration date from maternal report, vaccine
record, or medical record
Weeks’
gestation at
study
enrollment Timing of
study
enrollment,
Confounder Maternal rep ort
Medical record Weeks’ g estation at time of study enrollment,
continuous and categorical (<13, 13.1 -19.9,
>20),
Trimester of
Pfizer -
BioN Tech
COVID -19
vaccin e Timing of
exposure ,
Confounder Maternal report
Vaccine record (e.g.,
COVID -19 vaccine
or yellow card)
Medical record Gestational week of vaccination by trimester
catego ry:
1st trimester defined as 30 days prior to first day
of LMP through 13.0 weeks’ gestation
2nd trimester defined as 13.1 weeks ’ gestation
through 26.0 weeks’ gestation
3rd trimester defined as 26.1 weeks’ gestation to
the end of pregnancy
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236349
FDA-CBER-2022-5812-0236350
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 20 Table 2. Variables Related to Defining the Exposure, Comparator, Outcomes,
Exclusions, and Date/Time
Variable Role Data source(s) Operational definition
Spontaneous
abortion Pregnancy
outcome Maternal report
Medical record Non-deliberate embryonic or fetal death that
occurs prior to 20 weeks’ gestation from first day
of LMP (Prager et al., 2021 )
Stillbirth Pregnancy
outcome Maternal report
Medical record A non -deliberate fetal death that occurs at or
after 20 weeks’ gestation from first day of LMP
but prior to delivery (CDC, 2021 (b) )
Preterm
delivery Pregnancy
outcome Maternal report
Medical record A spontaneous or induced delivery at <37
gestational weeks from first day of LMP (CDC,
2021(c) )
Small for
gestational age Infant
outcome Maternal report
Medical record Birth size (weight, length, or head circumference)
≤10th percentile for sex and gestational age using
NCHS pediatric growth curves for full term
infants. Prenatal growth curves specific to
preterm infants are used for preterm infants
(Olsen et al., 2010 ); (Schlaudecker et al., 2017 )
Small for age
postnatal
growth at one
year of age Infant
outcome Medical record Postnatal size (weight, length or head
circumference) ≤10th percentile for sex and age
using NCHS pediatric growth curves and adjusted
postnatal age for preterm infants
Live birth Outcome -
specific
inclusion/
exclusion Maternal report
Medical record Singleton, twin or higher order gestation resulting
in at least one live born infant
Abbreviations: CDC, Centers for Disease Control and Prevention; LMP, last menstrual period; MACDP,
Metropolitan Atlanta Congenital Defects Program; NCHS, National Center for Health Statistics; OTIS,
Organization of Teratology Information Specialists; US, United States.
10.3. Classification of Pregnancy and Infant Outcomes
The following pregnancy and infant outcome variables (refer to Table 2) are obtained by
maternal report and/or medical record review (when available) as part of existing
procedures for the pregnancy registry.
10.3.1. Major Congenital Malformations
• Defined for pregnancies ending with at least one live born infant as a defect that has
either cosmetic or functional significance to the child (e.g., a cleft lip) diagnosed
during pregnancy up to 1 year of age as reported by participants/ healthcare
providers or identified throu gh medical record review.
• A pregnancy with multiple births is counted as one malformed outcome if one or
more of the infants/fetuses are malformed.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236351
FDA-CBER-2022-5812-0236352
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 22 1 year of age will be included if it is confirmed as a heart defect by a diagnostic
procedure such as cardiac ultrasound. Similarly, a midline cutaneous marker at L 2-
L3 noted will be included as occult spinal dysraphism only if confirmed by
appropriate imaging studies.
10.3.2. Spontaneous abortion
Defined as non -deliberate embryonic /fetal death (miscarriage) which occurs prior to 20
weeks’ gestation . In pregnancies involving multiples with one or more of the outcomes
ending in spontaneous abortion, when there are no live births, the pregnancy is counted as
one spontaneous abortion event; however, when the pregnancy ends in at least one live -born
infant, the pregnancy is count ed as a live birth outcome.
10.3.3. Elective termination/abortion
Defined as deliberate termination of pregnancy at any time in gestation. Reasons for elective
termination/abortion are captured and are classified as due to medical reasons or social
reasons.
10.3.4. Stillb irth
Defined as non -deliberate fetal death any time in gestation at or after 20 weeks ’ gestation . In
pregnancies involving multiples with one or more of the outcomes ending in stillbirth, when
there are no live births, the pregnancy is counted as one still birth event; however, when the
pregnancy ends in at least one live -born infant, the pregnancy is counted as a live birth .
10.3.5. Preterm delivery
Defined as live birth prior to 37 weeks’ gestation.
10.3.6. Small for gestational age at birth
Defined separately for weight, length, and head circumference (binary endpoints) as birth
size less than or equal to the 10th centile for sex and gestational age using National Center
for Health Statistics (NCHS) standard pediatric growth curves for full term or preterm
infants (CDC, 200 0). Prenatal growth curves specific to preterm infants will be used for
premature infants ( Olsen et al., 2010 ).
10.3.7. Small for age postnatal growth at one year of age
Defined separately for weight , length, and head circumference (binary endpoints) as
postnatal size less than or equal to the 10th centile for sex and age using NCHS pediatric
growth curves (CDC, 2000 (CDC, 2000) ,), and adjusted chronological age for preterm infants
if the postnatal measurement is obtained at less than 1 year of age (CDC, 2000 (CDC,
2000 ).). The measurements obtained closest to one year of age, and within three months prior
to or after one year of age (9 -15 months) are used.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236353
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 23 10.4. Demographic and Clinical Characteristics
Potential confounders and other covariates to be collected include maternal age,
race/ethnicity, socioeconomic status, pregnancy and health history, lifestyle factors,
comorbidities, medication, vaccine and vitamin/mineral exposures, and prenatal tests.
Table 3 provides a description of corresponding variables to be included in this study.
Table 3. Demographic and Clinical Variables
Variable Role Data source(s) Operational definition
Age Confounder Maternal report Maternal age (years) at due date, continuous
and categorical (<25, 25 -29, 30 -34, >34)
Race Confounder Maternal report Maternal/paternal race (Caucasian/White,
Black, Asian/Pacific Islander, Native
American, Other)
Ethnicity Confounder Maternal report Maternal/paternal ethnicity (Hispanic, Non -
Hispanic)
Education Confounder Maternal report Maternal e ducational category (years of
completed education <12, 12 -15, >15)
Socioeconomic
category Confounder Maternal report Hollingshead Socioeconomic Status (SES)
based on maternal and paternal occupation
and education (Categorical: high 1 -2,
moderate 3; low 4 -5) (Hollingshead, 1975 )
Geographic area of
residence Confounder Maternal report Geographic area of residence (US, Canada)
Referral source Confounder Maternal report Source options Sponsor, OTIS service, health
care provider (HCP), Internet, Referrals from
other University of California Medical
Centers , Other
Height Confounder Maternal report Maternal height in centimeters (cm)
Pre-pregnancy body
weight Confounder Maternal report
Medical record Maternal pre -pregnancy body weight in
kilograms (kg)
Pre-pregnancy body
mass index (BMI) Confounder Maternal report Calculated as maternal pre -pregnancy body
weight in kilograms (kg) divided by maternal
height in meters squared (m2). Categorized as:
Underweight = <18.5 kg/m2
Normal weight = 18.5-24.9 kg/m2
Overweight = 25-29.9 kg/m2
Obese = ≥30 kg/m2
Number of prior
pregnancies Confounder Maternal report
Medical record Number of times ever pregnant prior to
current pregnancy (1, 2-3, 4-5, ≥6)
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236354
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 24 Table 3. Demographic and Clinical Variables
Variable Role Data source(s) Operational definition
Number of previous
live birth or stillbirth
deliveries Confounder Maternal report
Medical record Number of previous live birth or stillbirth
deliveries , i.e., parity defined as number of
previous pregnancies ending in a live or
stillbirth after 24 weeks’ gestation (0, 1-2, 3-4,
≥5)
Previous pregnancies
with a major
congenital
malformation Confounder Maternal report ≥1 previous pregnancy with a major
structural or chromosomal defect diagnosed
in utero or post -partum – Yes/No
Type of major
congenital
malformation in
previous pregnancies Confounder Maternal report Specific major structural or chromosomal
defect in previous pregnancies that has either
cosmetic or functional significance to the
child (e.g., a cleft lip) classified using CDC
coding criteria
Number of previous
pregnancies ending in
spontaneous abortion Confounder Maternal report
Medical record Number of previous pregnancies ending in
spontaneous abortion (0, 1, 2, ≥3)
Number of previous
pregnancies ending in
elective
termination/abortion Confounder Maternal report
Medical record Number of previous pre gnancies ending in
elective termination/abortion (0, 1, 2, ≥3)
Previous pregnancies
ending in preterm
delivery Confounder Maternal report ≥1 previous pregnancy ending in preterm
delivery – Yes/No
Previous pregnancies
ending in fetal growth
restriction Confounder Maternal report ≥1 previous pregnancy ending in fetal growth
restriction – Yes/No
Number of previous
ectopic pregnancies Confounder Maternal report
Medical record Number of previous pregnancies ending in
ectopic pregnancy (0, 1, 2, ≥3); ectopic
pregnancy is defined as a pregnancy in which
the fetus develops outside the uterus, typically
in a fallopian tube
Family history of
genetic disorders and
major congenital
malformations Confounder Maternal report Any family history of a major structural or
chromosomal defect that has either cosmetic
or functional significance to the child as
defined using the CDC coding criteria –
Yes/No
Prenatal vitamin,
multivitamin, or folic
acid use in pregnancy Confounder Maternal report Prenatal, multivitamin, or folic acid
supplement use by timing (began prior to
conception, post -conception only, not taken at
all)
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236355
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 25 Table 3. Demographic and Clinical Variables
Variable Role Data source(s) Operational definition
Alcohol use in
pregnancy Confounder Maternal report Light or moderate alcohol use based on an
average dose and frequency . Participants are
classified in this group if they have had any
amount of alc ohol in pregnan cy but have not
exceeded >14 drinks per week for 4 or more
weeks after conception. – Yes/No (heavy
alcohol use is classified as a human teratogen
which is exclusionary, thus no participants in
the study would have heavy use)
Tobacco use in
pregnancy Confounder Maternal report Any tobacco use during pregnancy – Yes/No
Prenatal diagnostic
tests prior to study
enrollment Confounder Maternal report
Medical record ≥1 diagnostic tests performed during
pregnancy prior to study enrollment
(Ultrasound level 1, Ultrasound level 2,
Chorionic Villus Sampling, and
Amniocentesis)
Prenatal diagnostic
tests any time during
pregnancy Confounder Maternal report
Medical record ≥1 diagnostic tests performed any time in
pregnancy (Ultrasound level 1, Ultrasound
level 2, Chorionic Villus Sampling, and
Amniocentesis)
Pregnancy
complications Confounder
or Mediator Maternal report
Medical record Pregnancy induced hypertension – Yes/No
Preeclampsia – Yes/No
Gestational diabetes – Yes/No
Each pregnancy complication will be
considered as a confounder for the endpoints
of preterm delivery, small for gestational age
at birth, and small for age postnatal growth at
one year of age. If the complication occurs
after exposure to the Pfizer -BioNTech
COVID -19 vaccine in pregnancy , the
complication will also be evaluated as a
potential mediator of those same endpoints
Comorbid maternal
medical history Confounder Maternal report
Medical record Comorbid maternal medical history current
diagnosis in pregnancy
Chronic hypertension – Yes/No
Asthma – Yes/No
Psychiatric disorder – Yes/No
Immune disorders – Yes/No
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236356
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 26 Table 3. Demographic and Clinical Variables
Variable Role Data source(s) Operational definition
Current medication
use Confounder Maternal report Prescription and over -the-counter
medications are captured for the period of
time from the first day of LMP through the
end of pregnancy. Dose, frequency, duration,
and indication including stop and start dates
are collected ; specific categories of
medication treatments in pregnancy :
Antihypertensive medi cations – Yes/No
Asthma medications – Yes/No
Psychotherapeutic medications – Yes/No
Immune modulators – Yes/No
Exposure to the
influenza or Tdap
vaccine Confounder Maternal report
Medical record Exposure to an influenza or Tdap vaccine any
time from one month prior to first day of LMP
up to and including end of pregnancy
COVID -19 vaccine
prior to pregnancy Confounder Maternal report
Medical record Exposure to at least one dose of any COVID -
19 vaccine any time prior to one month before
the first day of LMP
COVID -19 symptoms
during pregnancy Confounder Maternal report Symptom inventory ( Annex 2: Table A -2)
COVID -19 infection
positive test during
pregnancy Confounder Maternal report
Medical record
Test results positive for COVID -19 (e.g., viral
RNA/PCR) from maternal report or medical
record
COVID -19 infection
positive test prior to
pregnancy Confounder Maternal report
Medical record Test results positive for COVID -19 (e.g., viral
RNA/PCR) from maternal repor t or medical
record
Abbreviations: BMI, body mass index; CDC, Centers for Disease Control and Prevention;
cm, centimeters; kg, kilograms; HCP, health care provider; LMP, last menstrual period;
OTIS, Organization of Teratology Information Specialists; PCR, polymerase chain reaction; RNA,
ribonucleic acid .
11. MISSING DATA
Multiple imputation (MI) will be conducted to handle the missing data. Missing values
typically occur in less than 5% of the cases for any single covariate based on prior
experience (). (Chambers, et al., 2019). They are assumed to be missing at random. When
there are missing values in any of the selected confounders, multiple imputation (MI) will be
conducted, using the R package MICE (multivariate imputation by chained equations ).) (van
Buuren S and Groothuis -Oudshoorn K , 2011) . Demographic information, pregnancy history,
information about current pregnancy, and concurrent diseases will be used as possible
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236357
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 27 ‘predictors’ in MICE. The generalized R2 will be used as a measure of correlations between
each ‘target’ (i.e., the variable to be imputed ) and a predictor. If the value of R (i.e., square
root of R2) is over 0.1, the variable will be retained as a predictor. Predictors with strong
collinearity might be excluded (for example, one of race and ethnicity, and one of gravidity
and parity). Altho ugh each outcome within each cohort comparison has its unique set of
participants, MI is conducted for missing data using the entire dataset, i.e., on all cohorts
combined.
For the outcome of spontaneous abortion, for some cases the exact date of the event might be
unknown, and instead a window for possible spontaneous abortion time is available. This is
known as interval censored data, and can also be handled using MI . (Pan, 2000) An exact
spontaneous abortion time will be imputed by sampling uniformly from the corresponding
time window.
The number of above imputations will be 10, i.e., 10 datasets with imputed data will be
created. Each imputed data set gives a point estimate of the regression coefficients as well as
its standard deviation, which will be co mbined across the 10 datasets to obtain the final
estimate of the causal RR/HR and their 95% CI’s . (Little and Ru bin, 2002) .
12. STATISTICAL METHODS AND DATA ANALYSIS
12.1. Data Cleaning and Preparation of Datasets
The original data will be cleaned and validated by the OTIS Data Manager and exported in
the form of multiple datasets to the study statisticians. A checklist will be filled out and
reviewed prior to the analysis to ensure the original data have been properly validated and
meet the study criteria. There wi ll be close communications between the statisticians and the
OTIS Research Manager during this process. Based on the original data, the statisticians
derive variables to be used for the analyses as described in the definitions of the variables in
Table 2 and Table 3.
12.2. Demographic and Baseline Characteristics
The distributions of demographic and baseline characteristics will be summarized within
each exposure cohort. Continuous variables will be summarized using the following
statistics: mean, standard deviation, minimum, 1st quartile, median, 3rd quartile, and
maximum. All categorical variables will be summarized using counts and percentages.
Missing data or unknown responses will not be counted in the perc entages.
12.3. Primary Analyses
The following measures will be calculated for the pregnancy and infant outcomes: birth
prevalence of major congenital malformations; incidence rates of spontaneous abortion,
elective termination/abortion, stillbirth, and preterm d elivery; and incidence proportions of
small for gestational age and small for age postnatal growth at one year of age. For each
outcome, risk estimates will be described separately for the Pfizer -BioNTech COVID -19
vaccine exposure cohort and the comparator cohort . The outcomes will be compared
between the Pfizer -BioNTech COVID -19 vaccine -exposed cohort and the comparator cohort.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236358
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 28 Where feasible, comparisons will also be made using methods to control potential
confounding.
The eligible populations and time pe riods for analyses may differ by pregnancy outcome
depending on a participant’s exposure status, timing of vaccine exposure/study enrollment as
it relates to the at -risk period for an outcome, in addition to type of birth ( Table 4 ). The
denominators for each exposure/comparator cohort and outcome analysis will include
participants as follows:
Table 4. Denominators for Outcomes by Exposure/Comparator Cohort
Outcome Pfizer -BioNTech COVID -19
Vaccine -Exposed
(Exposure) COVID -19 Vaccine -
Unexposed
(Comparator)
Major congenital
malformations Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP
through their first trimester and
whose pregnancy resulted in >1
live birth Participants who did not receive
any COVID -19 vaccines any
time from one month before the
first day of LMP up to and
including end of pregnancy and
whose pregnancy resulted in >1
live birth
Spontaneous
abortion Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP up
to 20 weeks’ gestation and were
enrolled in the study prior to 20
weeks’ gestation Participants who did not receive
any COVID -19 vaccines any
time from one month before the
first day of LMP up to and
including end of pregnancy and
were enrolled in the study prior
to 20 weeks’ gestation
Elective
termination/abortion Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP up
to and including end of
pregnancy Participants who did n ot receive
any COVID -19 any time from
one month before the first day of
LMP up to and including end of
pregnancy
Stillbirth Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP up Participants who did not receive
any COVID -19 vaccines any
time from one month before the
first day of LMP up to and
including end of pregnancy, but
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236359
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 29 Table 4. Denominators for Outcomes by Exposure/Comparator Cohort
Outcome Pfizer -BioNTech COVID -19
Vaccine -Exposed
(Exposure) COVID -19 Vaccine -
Unexposed
(Comparator)
to and including end of
pregnancy, but must have
follow -up at or after 20 weeks’
gestation must have follow -up at or after
20 weeks’ gestation
Preterm delivery Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP up
to 37 weeks’ gestati on, were
enrolled in the study prior to 37
weeks’ gestation, and whose
pregnancy resulted in a live -
born singleton Participants who did not receive
any COVID -19 vaccines any
time from one month before the
first day of LMP up to and
including end of pregnan cy,
were enrolled in the study prior
to 37 weeks’ gestation, and
whose pregnancy resulted in a
live-born singleton
Small for gestational
age Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP up
to and including end of
pregnancy and whose
pregnancy resulted in a live-
born singleton Participants who did not receive
any COVID -19 vaccines any
time from one month before the
first day of LMP up t o and
including end of pregnancy and
whose pregnancy resulted in a
live-born singleton
Small for age
postnatal growth at
one year of age Participants who received the
Pfizer -BioNTech COVID -19
vaccine (1st, 2nd, 3rd, or booster
dose) any time from one month
before the first day of LMP up
to and including end of
pregnancy and whose
pregnancy resulted in a live-
born singleton Participants who did not receive
any COVID -19 vaccines any
time from one month before the
first day of LMP up t o and
including end of pregnancy and
whose pregnancy resulted in a
live-born singleton
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236360
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 30 12.3.1. Major Congenital Malformation
For the endpoint of major congenital malformations , the exposure of interest is at least one
dose of the Pfizer -BioNTech COVID -19 vaccine during the first trimester. The primary
comparison will be the birth prevalence o f major congenital malformations between the
exposed cohort and the comparator cohort among pregnancies resulting in at least one live
born infant . The birth prevalence of ma jor congenital malformations will be calculated for
each cohort as the number of pregnancies ending in at least one liveborn infant with a major
congenital malformation divided by the number of all pregnancies for women in the cohort
ending in at least one live born infant (Table 4), multiplied by 100. A point estimate of the
crude (i.e., unadjusted) risk ratio (RR) of the exposed cohort versus the comparison cohort,
as well as its 95% confidence interval (CI), will be computed using normal approximation
method. When the expected frequency of any of the cells of the contingency table is less than
five, the CI will be obtained by an exact method using the software StatXact. The method is
based on inverting an unconditional exact hypothesis test (Agresti and Min, 2001 ).
Due to the observational nature of the study, the above crude estimate of RR will be adjusted
for potential confounders (Rosenbaum, 2002 ), provided that there are sufficient number of
events. A list of potential confounders will be provided in a separate appendix to the SAP for
each outcome prior to the final analysis, based on scientific knowledge including literature
review. In addition, all of the following three criteria will be applied in accordance with the
definition of confounders ( Greenland et al., 1999 ; Xu et al., 2018 ):
1) by assessi ng each considered variable in a logistic regression model containing the
exposure variable and the outcome variable to determine if inclusion of that single covariate
changes the estimate of the odds ratio (OR) for exposure by 10% or more: |OR 2 – OR 1|/OR 1
≥ 10% where OR 1 is the crude OR and OR 2 is the OR adjusting for the covariate in question;
2) standardized mean differences ( SMD) greater than 0.1;
3) association with the outcome with p -value < 0.2 in the unexposed cohort (using chi -square
test or two sample t -test). Care will be taken not to include those variables that are strongly
associated with the exposure variable but only weakly associated with the outcome variable
(e.g, instrument -like variables ) (Brookhart et al., 2006 ).
The confounders identified above will be used to build the propensity score for exposure
(Rosenbaum, 2002 ). R package ‘twang’ or similar R package available at the time of
analysis will be used for this purpose, follo wing which SMD will be used to check the
balance of the covariates between the cohorts.
The primary analysis will estimate the effect of exposure on the risk for major structural
defect as the causal risk ratio ( Hernán and Rob ins, 20 20) using inverse probability of
treatment weighting (IPTW) . The causal risk ratio is defined as 𝑃(𝑌𝑎=1=1)/P(𝑌𝑎=0=1),
where 𝑌𝑎=1 is the potential outcome under exposure and 𝑌𝑎=0 is the potential outcome
without exposure . The risks for the potential outcomes will be estimated using the inverse
probability weighting:
𝑃̂(𝑌𝑎=0=1)=𝑛−1∑𝑖=1𝑛𝑤𝑖(𝐴𝑖−1)𝑌𝑖(1− 𝐴𝑖)𝑌𝑖, 𝑃̂(𝑌𝑎=1=1)=𝑛−1∑𝑖=1𝑛𝑤𝑖𝐴𝑖𝑌𝑖,
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236361
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 31 𝑤here 𝑌𝑖 and 𝐴𝑖 denote the observed outcome ( 𝑌𝑖=1 or 0 for presence or absence of a major
congenital malformation) and exposure status ( 𝐴𝑖=1 or 0 for exposed or unexposed) among
the 𝑛 pregnancies, and 𝑤𝑖 is the inverse of the probability of being in the observed e xposure
group estimated using the propensity score, which has been stabilized and further trimmed to
be between 0.1 and 10 if necessary (Austin and Stuart, 2015 ). The causal risk ratio is then
estimated by 𝑃̂(𝑌𝑎=1=1)/𝑃̂(𝑌𝑎=0=1). The 95% CIs will be obtained using the bootstrap
estimated variance and asymptotic normality, with 200 bootstrap samples, i.e., resampling
with replacement on the pregnant women. The propensity score will be re -computed for each
bootstrap sample to account for variability in the estimation of the weights.
Provided there are sufficient number of events, i.e., at least 10 events per parameter in the
regression model, an additional analysis will be conducted using outcome regression (Xu et
al., 2018 ; Vansteelandt and Daniel, 2014 ). A logistic regression model will be fitted with
major congenital malformation (Y) as the outcome, and exposu re (A) and propensity score
(L) as regressors.
Outcomes regression tends to be more stable and efficient than IP TW especially given the
expected rare number of events (Xu et al., 2018 (Xu et al., 2018 ),), and has known robust
properties against model misspecification (Vansteelandt and Daniel, 2014 (Vansteelandt and
Daniel, 2014 ).). In a second step , standardization will be performed to obtain the estimated
causal risk ratio (Hernán and Robins, 2020 ), which has the interpretation as the marginal or
population averaged risk ratio. Note that
𝑃(𝑌𝑎=1=1)= ∑𝑃(𝑌=1|𝐴=1,𝐿=𝑙)𝑃(𝐿=𝑙) 𝑙 .
The above can be estimated by first predicting the potential outcomes 𝑃̂(𝑌𝑎=1=
1|𝐴=1,𝐿=𝑙)=𝑃̂(𝑌=1|𝐴=1,𝐿=𝑙), using the fitted outcomes regression model for
the whole sample (i.e., both exposed and unexposed), assuming that their treatments are all a
= 1. 𝑃̂(𝑌𝑎=1=1) will be estimated by averaging these predicted values over the distribution
of L, as in the equation above. P(𝑌𝑎=0=1) is estimated in a similar way. Finally, the causal
risk ratio will be estimated by dividing these two estimated probabilities . The CI’s are
obtained by bootstrap with 10,000 200 bootstrap samples, i.e. resampling with replacement of
the pregnant women.
12.3.2. Spontaneo us Abortion, Stillbirth, and Preterm Delivery
The analyses of spontaneous abortion, stillbirth, and preterm delivery are complicated by
several factors based on the timing of study entry and the at -risk period for the outcome. The
eligible populations and time periods for the analyses of each of these outcomes are as
follows:
• Spontaneous abortion: Only those women who are enrolled prior to 20.0 weeks of
gestation in both exposed and comparator cohorts, and only those women exposed to
at least one dose of th e Pfizer -BioNTech COVID -9 vaccine from 30 days before the
first day of LMP up to 20.0 weeks’ gestation.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236362
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 32 • Stillbirth: Only those pregnancies that reach at least 20 weeks of gestation are eligible
for analysis.
• Preterm delivery: Only those women exposed to at least one dose of the Pfizer -
BioNTech COVID -19 vaccine, or enrolled in the study for unvaccinated comparator,
prior to 37 weeks’ gestation and whose pregnancies resulted in a live -born singleton;
pregnancies with twins or higher order multiples will be ex cluded.
Because women can enter the study at arbitrary times in gestation (i.e., left truncation ),
participants are not followed from gestational age zero. Those who experience the event
prior to enrollment will never enter the study, leading to left truncation of the time to event.
This can produce selection bias if early vaccination time causes early time to event. The
event time is also right -censored whenever a subject is lost to follow -up prior to observing
the outcome. In addition, vaccine exposu re can occur prior to or throughout enrollment, so
exposure status is time -dependent . In order to address these issues, survival analysis methods
will be used to handle possible left truncation, right -censoring, and time -dependent exposure
or confounding. The Cox proportional hazards marginal structural model (MSM)
incorporating time -dependent vaccine exposure and relevant covariates will be used to
estimate the causal hazard ratio (HR) and 95% CIs for exposure to the Pfizer -BioNTech
COVID -19 vaccine (Hernán et al, 2001). For 𝑡≥0 let 𝑎(𝑡) be an indicator for exposure to
the Pfizer -BioNTech COVID -19 vaccine by time 𝑡 during the exposure window and let 𝑇𝑎
denote the potential time to spontaneous abortion had a subject followe d the exposure profile
𝑎. The marginal structural model is:
𝜆𝑎(𝑡)=𝜆0(𝑡)exp(𝛽𝑎(𝑡))
where 𝜆𝑎(𝑡) is the hazard of 𝑇𝑎 at time 𝑡, 𝜆0 is the baseline hazard function representing the
risk for a subject who never receives the vaccine, and exp(𝛽) is the causal relative risk for
the effect of exposure.
The method that will be used to estimate the MSM depends on whether there are time -
dependent confounders that are also affected by previous exposure to the Pfizer -BioNTech
COVID -19 vaccine (Hernán et al, 2001). This study considers two potential sources of time -
dependent confounding: COVID -19 circulation season and COVID -19 infection in
pregnancy prior to vaccination . Circulation season is an external mechanism that is not
affected by a subject’s previous vaccine exposure . On the other hand, infection during
pregnancy is likely to be affected by previous vaccine exposure . The estimation method
therefore depends on whether infection is included as a confounder .
If infection during pregnancy is not selected as a confounder, the MSM can be consistently
estimated using multivariable regression. In this approach, a t ime-dependent proportional
hazards model for the time to spontaneous abortion will be fit, with effects for the time -
varying exposure as well as all selected confounders. The usual 95% Wald confidence
intervals will be computed using the standard software.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236363
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 33 If infection during pregnancy is selected as a confounder, the MSM can be consistently
estimated using IPTW . To obtain the weights, a proportional hazards model is fit for the time
to vaccine exposure during the exposure window, adjusting for the selected confounders (Xu
et al., 2014 ; Yang et al., 2018 .). Estimates of the conditional survival function 𝑆(𝑡|𝑙) and
density function 𝑓(𝑡|𝑙) given the confounders 𝑙 are then used to compute time -dependent
propensity scores and weights . At a fixed time 𝑡, the weight for a woman with confounders
𝐿(𝑡) who is still at risk for SAB by time 𝑡 is 𝑤(𝑡)=1/𝑃𝑆(𝑡), where 𝑃𝑆(𝑡) is computed as
follows:
1) For a subject who is not exposed during the wind ow,
𝑃𝑆(𝑡)=𝑆(𝑡|𝐿(𝑡)) if they have not been fully vaccinated prior to the window
𝑃𝑆(𝑡)=1 if they have been fully vaccinated prior to the window
2) For a subject who is exposed during the window at time 𝐷,
𝑃𝑆(𝑡)={𝑆(𝑡|𝐿(𝑡)) if 𝑡<𝐷
𝑓(𝐷|𝐿(𝐷) if 𝑡≥𝐷
When there is a strong association between confounders and exposure time, the weights can
be large for some subjects, leading to a large variance in the IPTW estimator . Variability can
be reduced using stabilization and trimming ( Austin and Stuart, 2015 ). The weights will be
stabilized by multiplying by 𝑃𝑆0(𝑡), which is computed analogous to 𝑃𝑆(𝑡) except using the
baseline quantities 𝑆0(𝑡)=𝑆(𝑡|0) and 𝑓0(𝑡)=𝑓(𝑡|0).) (Yang et al., 2018) If necessary, the
weights will be further trimmed to lie between 0.1 and 10.
The exposure time is not observed for women who never enter the study due to left
truncation of the outcome . As a consequence, the exposure time may also be subject to left
truncation. This is a p otential source of bias in the PS estimates, and a possible limitation of
the IPTW method.
The standard error se(𝛽̂) of the causal exposure effect will be estimated using the
nonparametric bootstrap, and 95% CIs will be computed as 𝛽̂±1.96 se(𝛽̂).
12.3.3. Electiv e Termination/Abortion
The analysis of elective termination/abortion will be descriptive as the number of events is
expected to be low.
12.3.4. Small for Gestational Age at Birth and Small for Age Postnatal Growth at One
Year of Age
The following are binary endpo ints: small for gestational age at birth in weight, length, and
head circumference; and small for age postnatal growth at one year of age 10th centile in
weight, length, and head circumference. The analysis of the incidence proportion of infants
with each of these outcomes will be similar to the analysis of the outcome major congenital
malformations (see Section 10.3.1 Section 10.3.1. . “Major Congenital Malformation) and will
be restricted to pregnancies ending in a live born singleton ; pregnancies with twins or higher
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236364
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 34 order multiples will be excluded . For these outcomes, exposure to any dose of the Pfizer -
BioNTech COVID -19 vaccine in any trimester will be compared to the unexposed cohort.
12.4. Secondary Analyses
12.4.1. Stratified/Subgroup Analyses
For the endpoint of major congenital malformations, the comparison will also be carried out
within each of two strata, according to whether the participant had prenatal diagnostic
testing, such as level 2 ultrasound, amniocentesis or chorionic villus sampling, prior to
enrollment in the study or not.
In stratified analyses, the separate effects of vaccine exposure during the first, second, and
third trimester of pregnancy compared to unexposed will be studied as applicable to the
following pregnancy outcomes: preterm delivery, small for gestat ional age, and small for age
postnatal growth at one year of age.
A variable for prior COVID -19 infection will be used in the propensity score generation for
the adjusted analyses, and, if numbers permit, a stratified analysis for prior COVID -19
infection will be conducted as well.
12.4.2. Individual Dose Effects
We will address the potentially differential effects of the 1st, 2nd, 3rd, and booster doses of the
Pfizer -BioNTech COVID -19 vaccine during the relevant pregnancy window . Each dose will
be considered a sep arate exposure, and a marginal structural model (Robins et al., 2000 ;
Hernán et al., 2001 ) with independent effects for the exposures will be evaluated for each
pregnancy outcome.
Binary Outcomes : For an exposure window of interest, let 𝐴1, 𝐴2, 𝐴3, and 𝐴4 be indicators
for exposure to the 1st, 2nd, 3rd, and booster dose s, respectively, during the window; i.e. 𝐴𝑖=
1 if the 𝑖th dose occurs during the window and 𝐴𝑖=0 otherwise, for 𝑖= 1, 2, 3 and 4 . The
sequence (𝐴1,𝐴2,𝐴3,𝐴4) describes the pattern of exposures during the window . Let
𝑌𝑎1,𝑎2,𝑎3,𝑎4 be the potential outcome that would occur if a woman was subject to the
exposures 𝐴1=𝑎1,𝐴2=𝑎2,𝐴3=𝑎3 and 𝐴4=𝑎4, where each of 𝑎1,𝑎2,𝑎3 and 𝑎4 are 0 or
1. The linear logistic MSM with additive effects for each vaccine dose assumes the fo rm
logit 𝑃(𝑌𝑎1,𝑎2,𝑎3,𝑎4=1)= 𝛽0+𝛽1𝑎1+𝛽2𝑎2+𝛽3𝑎3+𝛽4𝑎4.
In this model, 𝛽𝑖 is the log OR representing the causal effect of exposure to the 𝑖th dose. The
causal effects are estimated from the observed data using an IPTW estimator.
The weigh ts for the IPTW estimator will be constructed from regression models for each of
the exposure doses 𝐴1, 𝐴2, 𝐴3, and 𝐴4, adjusted for the selected confounders 𝐿. Let 𝜋1(𝑏) be
the probability of receiving the 1st dose during the exposure window conditional on 𝐿 among
subjects that received 𝑏 doses prior to the expos ure window . Let 𝜋2(𝑏,𝑎1) be the probability
of receiving the 2nd dose during the exposure window conditional on 𝐿 among those subjects
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236365
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 35 with 𝐴1=𝑎1 and who received 𝑏 doses prior to the start of the window . Likewise, let
𝜋3(𝑏,𝑎1,𝑎2) be the proba bility of receiving the 3rd dose during the exposure window
conditional on 𝐿 among those subjects with 𝐴1=𝑎1 and 𝐴2=𝑎2 and who received 𝑏 doses
prior to the start of the window . Finally, let 𝜋4(𝑏,𝑎1,𝑎2,𝑎3) be the probability of receiving
the 4th dose during the exposure window conditional on 𝐿 among those subjects with 𝐴1=
𝑎1, 𝐴2=𝑎2 and 𝐴3=𝑎3 and who received 𝑏 doses prior to the start of the window . The
weight for a subject with exposures 𝐴1, 𝐴2, 𝐴3, and 𝐴4 and confounders 𝐿 is 𝑤=1/𝑃𝑆,
where 𝑃𝑆 can be decomposed in terms of 𝜋1(𝑏),𝜋2(𝑏,𝑎1),𝜋3(𝑏,𝑎1,𝑎2), and
𝜋4(𝑏,𝑎1,𝑎2,𝑎3) as follows:
1) 𝐴1=0, 𝐴2=0,𝐴3=0, and 𝐴4=0:
𝑃𝑆=1−𝜋1(0) if they have received no doses prior to the window
𝑃𝑆=1−𝜋2(1,0) if they have received one dose prior to the window
𝑃𝑆=1−𝜋3(2,0,0) if they have received two doses prior to the window
𝑃𝑆=1−𝜋4(3,0,0,0) if they have received three doses prior to the window
𝑃𝑆=1 if they have received four doses prior to the window
2) 𝐴1=1, 𝐴2=0,𝐴3=0, and 𝐴4=0: 𝑃𝑆=𝜋1(0)(1−𝜋2(0,1))
3) 𝐴1=0, 𝐴2=1,𝐴3=0, and 𝐴4=0: 𝑃𝑆=𝜋2(1,0)(1−𝜋3(1,0,1))
4) 𝐴1=0, 𝐴2=0,𝐴3=1, and 𝐴4=0: 𝑃𝑆=𝜋3(2,0,0)(1−𝜋4(2,0,0,1))
5) 𝐴1=0, 𝐴2=0,𝐴3=0, and 𝐴4=1: 𝑃𝑆=𝜋3(3,0,0,0)
6) 𝐴1=1, 𝐴2=1,𝐴3=0, and 𝐴4=0: 𝑃𝑆=𝜋1(0)𝜋2(0,1)(1−𝜋3(0,1,1))
7) 𝐴1=0, 𝐴2=1,𝐴3=1, and 𝐴4=0: 𝑃𝑆=𝜋2(1,0)𝜋3(1,0,1)(1−𝜋4(1,0,1,1))
8) 𝐴1=0, 𝐴2=1,𝐴3=0, and 𝐴4=1: 𝑃𝑆=𝜋2(1,0)(1−𝜋3(1,0,1))𝜋4(1,0,1,0))
9) 𝐴1=0, 𝐴2=0,𝐴3=1, and 𝐴4=1: 𝑃𝑆=𝜋3(2,0,0)𝜋4(2,0,0,1)
10) 𝐴1=1, 𝐴2=1,𝐴3=1, and 𝐴4=0: 𝑃𝑆=𝜋1(0)𝜋2(0,1)𝜋3(0,1,1)(1−
𝜋4(0,1,1,1))
11) 𝐴1=1, 𝐴2=1,𝐴3=0, and 𝐴4=1: 𝑃𝑆=𝜋1(0)𝜋2(0,1)(1−
𝜋3(0,1,1))𝜋4(0,1,1,0)
12) 𝐴1=0, 𝐴2=1,𝐴3=1, and 𝐴4=1: 𝑃𝑆=𝜋2(1,0)𝜋3(1,0,1)𝜋4(1,0,1,1)
13) 𝐴1=1, 𝐴2=1,𝐴3=1, and 𝐴4=1: 𝑃𝑆=𝜋1(0)𝜋2(0,1)𝜋3(0,1,1) 𝜋4(0,1,1,1)
Models for the 𝜋𝑖 terms will be estimated using the twang package in R as described in the
primary analysis . Each model will be fit using the subset of data indicated by the arguments
𝑏, 𝑎1,𝑎2 and 𝑎3; for example, the model for 𝜋3(0,1,1) will be fit using the selected
confounders in 𝐿 as predictors among subjects who received no doses prior to the window
(𝑏=0) and the 1st and 2nd dose during the window ( 𝑎1=1 and 𝑎2=1). The weights will
be stabilized and trimmed (Austin and Stuart, 2015 ).
Spontaneous Abortion and Preterm Delivery : For the survival outcomes, a proportional
hazards marginal structural model (MSM) with time -dependent covariates will be used to
estimate the causal hazard ratio for each of the exposure doses . Similar to the primary
analyses, let 𝑎1(𝑡), 𝑎2(𝑡), 𝑎3(𝑡), and 𝑎4(𝑡) be indicators for exposure duri ng the relevant
exposure window to the 1st, 2nd, 3rd and booster doses of the Pfizer -BioNTech COVID -19
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236366
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 36 vaccine by time 𝑡, and let 𝑇𝑎1,𝑎2,𝑎3,𝑎4 denote the potential time to SAB had a subject followed
the exposure functions 𝑎1, 𝑎2, 𝑎3, and 𝑎4. The marginal structural model is
𝜆𝑎1,𝑎2,𝑎3,𝑎4(𝑡)=𝜆0(𝑡)exp(𝛽1𝑎1(𝑡)+𝛽2𝑎2(𝑡)+𝛽3𝑎3(𝑡)+𝛽4𝑎4(𝑡))
where 𝜆𝑎1,𝑎2,𝑎3,𝑎4(𝑡) is the hazard of 𝑇𝑎1,𝑎2,𝑎3,𝑎4 at time 𝑡, 𝜆0 is the baseline hazard function
representing the risk for a subject who never receives either dose during the window, and
exp(𝛽𝑖) is the causal relative risks for the effect of exposure to the 𝑖th dose during the
window.
As in the primary analyses of the survival outcomes, the MSM can be estimated by either
multivariable regression or IPTW depending on whether prior infection is included as a time -
dependent confounder.
The multivariable regression model will be fit as in the primary analyses except that it will
include effects for the time -dependent exposures 𝐴1(𝑡), 𝐴2(𝑡), 𝐴3(𝑡), and 𝐴4(𝑡), which
indicate exposure by time 𝑡 to the 1st, 2nd, 3rd, and booster dose s during the exposure window,
respectively.
The IPTW method is based on PS models for the time to the 1st, 2nd, 3rd and booster dose s.
Similar to the primary analysis of the survival outcomes, proportional hazards models will be
fit for the time to each of the four doses during the exposure window, adjusting for the
selected confounders . The models yield estimates of the conditional survival function 𝑆𝑖(𝑡|𝑙)
and density function 𝑓𝑖(𝑡|𝑙) given the confounders 𝐿=𝑙 at time 𝑡, for each of the four dos es
(𝑖=1,2, 3, and 4). These estimates are used to compute time -dependent PS and weights . For
a fixed time 𝑡, consider a woman with confounders 𝐿(𝑡) who is still at risk for the outcome
by time 𝑡; let 𝐷1, 𝐷2,𝐷3, and 𝐷4 denote the time of exposure t o the 1st, 2nd, 3rd, and booster
doses, respectively, if they occur during the window . For this subject, the weight at time 𝑡 is
𝑤(𝑡)=1/𝑃𝑆(𝑡) where 𝑃𝑆(𝑡) is computed as follows:
1) For a subject with 𝐴1=0, 𝐴2=0,𝐴3=0, and 𝐴4=0,
𝑃𝑆(𝑡)=𝑆1(𝑡|𝐿(𝑡)) if they have received no doses prior to the window
𝑃𝑆(𝑡)=𝑆2(𝑡|𝐿(𝑡)) if they have received one dose prior to the window
𝑃𝑆(𝑡)=𝑆3(𝑡|𝐿(𝑡)) if they have received two doses prior to the window
𝑃𝑆(𝑡)=𝑆4(𝑡|𝐿(𝑡)) if they have received thre e doses prior to the window
𝑃𝑆(𝑡)=1 if they have received four doses prior to the window
2) For a subject with 𝐴1=1, 𝐴2=0, 𝐴3=0, and 𝐴4=0,
𝑃𝑆(𝑡)={𝑆1(𝑡|𝐿(𝑡)) if 𝑡<𝐷1
𝑓1(𝐷1|𝐿(𝐷1))𝑆2(𝑡|𝐿(𝑡)) if 𝑡≥𝐷1
3) For a subject with 𝐴1=0, 𝐴2=1, 𝐴3=0, and 𝐴4=0,
𝑃𝑆(𝑡)={𝑆2(𝑡|𝐿(𝑡)) if 𝑡<𝐷2
𝑓2(𝐷2|𝐿(𝐷2))𝑆3(𝑡|𝐿(𝑡)) if 𝑡≥𝐷2
4) For a subject with 𝐴1=0, 𝐴2=0, 𝐴3=1, and 𝐴4=0,
𝑃𝑆(𝑡)={𝑆3(𝑡|𝐿(𝑡)) if 𝑡<𝐷3
𝑓3(𝐷3|𝐿(𝐷3))𝑆4(𝑡|𝐿(𝑡)) if 𝑡≥𝐷3
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236367
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 37 5) For a subject with 𝐴1=0, 𝐴2=0, 𝐴3=0, and 𝐴4=1,
𝑃𝑆(𝑡)={𝑆4(𝑡|𝐿(𝑡)) if 𝑡<𝐷4
𝑓4(𝐷4|𝐿(𝐷4)) if 𝑡≥𝐷4
6) For a subject with 𝐴1=1, 𝐴2=1, 𝐴3=0, and 𝐴4=0,
𝑃𝑆(𝑡)={𝑆1(𝑡|𝐿(𝑡)) if 𝑡<𝐷1
𝑓1(𝐷1|𝐿(𝐷1))𝑆2(𝑡|𝐿(𝑡)) if 𝐷1≤𝑡<𝐷2
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑆3(𝑡|𝐿(𝑡)) if 𝑡≥𝐷2
7) For a subject with 𝐴1=0, 𝐴2=1, 𝐴3=1, and 𝐴4=0,
𝑃𝑆(𝑡)={𝑆2(𝑡|𝐿(𝑡)) if 𝑡<𝐷2
𝑓2(𝐷2|𝐿(𝐷2))𝑆3(𝑡|𝐿(𝑡)) if 𝐷2≤𝑡<𝐷3
𝑓2(𝐷2|𝐿(𝐷2))𝑓3(𝐷3|𝐿(𝐷3))𝑆4(𝑡|𝐿(𝑡)) if 𝑡≥𝐷3
8) For a subject with 𝐴1=0, 𝐴2=1, 𝐴3=0, and 𝐴4=1,
𝑃𝑆(𝑡)={𝑆2(𝑡|𝐿(𝑡)) if 𝑡<𝐷2
𝑓2(𝐷2|𝐿(𝐷2))𝑆4(𝑡|𝐿(𝑡)) if 𝐷2≤𝑡<𝐷4
𝑓2(𝐷2|𝐿(𝐷2))𝑓4(𝐷4|𝐿(𝐷4)) if 𝑡≥𝐷4
9) For a subject with 𝐴1=0, 𝐴2=0, 𝐴3=1, and 𝐴4=1,
𝑃𝑆(𝑡)={𝑆3(𝑡|𝐿(𝑡)) if 𝑡<𝐷3
𝑓3(𝐷3|𝐿(𝐷3))𝑆4(𝑡|𝐿(𝑡)) if 𝐷3≤𝑡<𝐷4
𝑓3(𝐷3|𝐿(𝐷3))𝑓4(𝐷4|𝐿(𝐷4)) if 𝑡≥𝐷3
10) For a subject with 𝐴1=1, 𝐴2=1, 𝐴3=1, and 𝐴4=0,
𝑃𝑆(𝑡)=
{ 𝑆1(𝑡|𝐿(𝑡)) if 𝑡<𝐷1
𝑓1(𝐷1|𝐿(𝐷1))𝑆2(𝑡|𝐿(𝑡)) if 𝐷1≤𝑡<𝐷2
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑆3(𝑡|𝐿(𝑡)) if 𝐷2≤𝑡<𝐷3
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑓3(𝐷3|𝐿(𝐷3))𝑆4(𝑡|𝐿(𝑡)) if 𝑡≥𝐷3
11) For a subject with 𝐴1=1, 𝐴2=1, 𝐴3=0, and 𝐴4=1,
𝑃𝑆(𝑡)=
{ 𝑆1(𝑡|𝐿(𝑡)) if 𝑡<𝐷1
𝑓1(𝐷1|𝐿(𝐷1))𝑆2(𝑡|𝐿(𝑡)) if 𝐷1≤𝑡<𝐷2
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑆4(𝑡|𝐿(𝑡)) if 𝐷2≤𝑡<𝐷4
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑓4(𝐷4|𝐿(𝐷4)) if 𝑡≥𝐷4
12) For a subject with 𝐴1=0, 𝐴2=1, 𝐴3=1, and 𝐴4=1,
𝑃𝑆(𝑡)=
{ 𝑆2(𝑡|𝐿(𝑡)) if 𝑡<𝐷2
𝑓2(𝐷2|𝐿(𝐷2))𝑆3(𝑡|𝐿(𝑡)) if 𝐷2≤𝑡<𝐷3
𝑓2(𝐷2|𝐿(𝐷2))𝑓3(𝐷3|𝐿(𝐷3))𝑆4(𝑡|𝐿(𝑡)) if 𝐷3≤𝑡<𝐷4
𝑓2(𝐷2|𝐿(𝐷2))𝑓3(𝐷3|𝐿(𝐷3))𝑓4(𝐷4|𝐿(𝐷4)) if 𝑡≥𝐷4
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236368
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 38 13) For a subject with 𝐴1=1, 𝐴2=1, 𝐴3=1, and 𝐴4=1,
𝑃𝑆(𝑡)
=
{ 𝑆1(𝑡|𝐿(𝑡)) if 𝑡<𝐷1
𝑓1(𝐷1|𝐿(𝐷1))𝑆2(𝑡|𝐿(𝑡)) if 𝐷1≤𝑡<𝐷2
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑆3(𝑡|𝐿(𝑡)) if 𝐷2≤𝑡<𝐷3
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑓3(𝐷3|𝐿(𝐷3))𝑆4(𝑡|𝐿(𝑡)) if 𝐷3≤𝑡<𝐷4
𝑓1(𝐷1|𝐿(𝐷1))𝑓2(𝐷2|𝐿(𝐷2))𝑓3(𝐷3|𝐿(𝐷3))𝑓4(𝐷4|𝐿(𝐷4)) if 𝑡≥𝐷4
The PS model for the 1st dose will be fit using data from subjects who did not receive any
doses prior to the exposure window. For the 2nd dose, the model will be fit using data from
subjects who received at one dose prior to or during the window . Similarly, for the 3rd and
boost er dose s, the model s will be fit using data from subjects who received all previous doses
prior to or during the window. Weight stabilization and trimming will be administered as in
the primary analyses.
Confidence intervals for either the multiple regres sion or IPTW method will be computed as
in the primary analyses .
12.4.3. Evaluation for a Pattern of Major Congenital Malformations
As most known human teratogens are associated with specific clusters of major congenital
malformations rather than an overall incre ase in the risk of all specific major congenital
malformations, a qualitative review will be conducted (Alwan and Chambers, 2015 ). The
following steps will be taken to evaluate any pattern of major congenital malformations :
A review of major congenital malformations will be made by category , including multiple
malformation syndromes, categories of major congenital malformations, such as cardiac
malformations, limb reduction anomalies, etc. A review of specific major congenita l
malformations will be conducted taking into consideration timing and biological plausibility.
In addition, specific defects that are plausibly related to second or third trimester exposure
will also be evaluated .
12.4.4. Lost to Follow -Up
Pregnancies enrolled in the cohort study for which outcome information is unobtainable
within 1 year after the estimated date of delivery are considered lost to follow -up. It is
possible that outcomes among pregnancies lost to -follow -up could differ from those with
documented ou tcomes. Because of differences in follow -up and reporting patterns, it is
currently not possible to assess with any certainty what impact with regard to potential
biases the lost to follow -up may have on any analysis of the cohort study. Should lost to
follow-up rates exceed 10%, in an attempt to address this potential source of bias , baseline
demographic characteristics including maternal age, socioeconomic status, race and ethnicity
will be compared between those lost -to-follow -up in each cohort vs. those who are retained.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236369
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 39 However, the OTIS Research Center prior experience has been that the lost to follow -up rate
is extremely low, typically 5% .
12.5. Sensitivity Analyses
The following sensitivity analyses will be performed for the outcome of major congenital
malformations.
• A sensitivity analysis will be performed for the outcome of major congenital
malformations among all pregnancies excluding those that are lost -to-follow -up. The
analysis population will be those with exposure to the Pfizer -BioNTech COVID -19
vaccine at any time from one month prior to the first day of LMP to the end of the
first trimester excluding those that are lost -to-follow -up. The analysis population for
the comparison cohort will be all pregnancies excluding those that are lost -to-follow -
up. The purpose of this analysis is to account for major congenital malformations that
may occur in pregnancies that are terminated or spontaneously lost and are therefore
excluded from the primary analysis of major congenital malformations among live
births .
• A second sensitivity analysis will be performed for the outcome of major congenital
malformations stratified on any abnormal finding (yes/no) among those with prenatal
testing prior to enrollment.
• In a third sensitivity analysis, both the exposure and co mparator cohorts will be
restricted to women who enrolled in the study during their first trimester.
• A fourth sensitivity analysis will be conducted to expand the comparator cohort to
include women who received any dose of the Pfizer -BioNTech COVID -19 vacc ine
only in their second or third trimester.
• Analys is using graphical presentation based on gestational timing of exposure to the
Pfizer -BioNTech COVID -19 vaccine will also be performed.
An additional sensitivity analysis will be performed for the outcome of preterm delivery
stratified on elective cesarian section or labor induction leading to delivery prior to 37
weeks’ gestation vs. delivery prior to 37 weeks’ gestation following spontaneous labor.
12.6. Interim Analyses
Descriptive analyses will be pre sented in annual interim reports, but no formal interim
statistical analysis is planned. The rationale for this is that given sample size limitations, an
interim analysis will likely be statistically underpowered to provide informative results. Each
report will be a composite of the cumulative data to date and will supersede any previous
reports. At each interim Scientific Advisory Board review, consideration will be given to any
findings that might indicate that a formal interim analysis should be performe d. The final
analysis will be conducted when the cohort study has been completed.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236370
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 40 12.7. Analysis Software
All summaries and statistical analyses will be performed using the current version of open -
source statistical programming language R and StatXact.
13. REFERENCE S
1. ACOG Practice Bulletin 700 (2017). Available at https://www.acog.org/clinical/clinical -
guidance/committee -opinion/articles/2017/05/methods -for-estimating -the-due-date.
Accessed November 18, 2021.
2. Agresti A, Min Y, (2001). On small -sample confidence inter vals for parameters in
discrete distributions. Biometrics 57, 963 –71.
3. Alwan S , Chambers CD (2015). Identifying human teratogens: an update. J Pediatr
Genet; 4:39 -41.
4. Andrews EA, Avorn J, Bortnichak EA, Chen R, Dai WS, Dieck GS, et al (1996).
Guidelines for Good Epidemiology Practices for drug, device, and vaccine research in
the United States. Pharmacoepidemiol Drug Saf; 5:333 8.
5.4. Austin PC and Stuart E (2015). Moving towards best practice when using inverse
probability of treatment weighting (IPTW) using the propensity score to estimate causal
treatment effects in observational studies. Stat Med;34:3661 -79.
6.5. Avalos LA, Galindo C, Li D -K (2012). A systematic review to calculate background
miscarriage rates using life tab le analysis. Birth Defects Res A Clin Mol Teratol
94:417 –23.
7.6. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, and Sturmer T (2006).
Variable selection for propensity score models. Am J Epidemiol;163:1149 -56.
8.7. CDC (2000). National Center for Heal th Statistics Individual Growth Charts.
https://www.cdc.gov/growthcharts/charts htm. Accessed November 23, 2021.
9.8. CDC (2017). Metropolitan Atlanta Congenital Defects Program (MACDP).
https://www.cdc.gov/ncbddd/birthdefects/macdp html. Accessed November 23, 2021.
10.9. CDC (2021a). V -Safe COVID -19 Vaccine Pregnancy Registry. Available at:
https://www.cdc.gov/coronavirus/2019 -
ncov/vaccines/safety/vsafepregnancyregistry html. Accessed October 30, 2021.
11.10. CDC (2021b). What is Stillbirth? Available at:
https://www.cdc.go v/ncbddd/stillbirth/facts html. Accessed November 18, 2021.
12.11. CDC (2021c). Premature Birth. Available at:
https://www.cdc.gov/reproductivehealth/features/premature -birth/index html. Accessed
November 18, 2021.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236371
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 41 12. Chambers CD, Johnson DL, Xu R. (2019). Birth out comes in women who have taken
Adalimumab in pregnancy: a prospective cohort study. PLOS ONE. 14. E0223603.
10.1371/JOURNAL.PONE.0223603.
13. Chambers CD, Johnson DL, Xu R, Luo YJ, Louik C, Mitchell AA, et al. (2016). Safety
of the 2010 -11, 2011 -12, 2012 -13, and 2013 -14 seasonal influenza vaccines in
pregnancy: Birth defects, spontaneous abortion, preterm delivery, and small for
gestational age infants, a study from the cohort arm of VAMPSS. Vaccine; 34:4443 -9.
14. Chambers CD, Braddock SR, Briggs GG, Einarson A, Johnson YR, Miller RK, et al.
(2001). Postmarketing surveillance for human teratogenicity: a model approach.
Teratology;64 :252 61.
15.14. FDA (2019). Draft Postapproval Pregnancy Safety Studies; Guidance for Industry.
Maryland, May 2019.
16.15. FDA (2021a). Pfizer COVID -19 Vaccine EUA Letter of Authorization reissued 02 -
25-21. Available at: https://www fda.gov/media/144412/download. Accessed March 3,
2021.
17.16. FDA (2021b). Comirnaty Approval Letter – August 23, 2021 . Available at:
https://www.fda.gov/media/151710/download . Accessed October 30, 2021.
18.17. FDA (2021c). Vaccine information fact sheet for recipients and caregivers about
Comirnaty (COVID -19 vaccine, mRNA) and the Pfizer -BioNTech COVID -19 vaccine
to prevent coronavirus disease 2019 (COVID -19) for use in individuals 12 years of age
and older. Available at: https://www.fda.gov/media/153716/download . Accessed
November 14, 2021 .
19.18. FDA (2021d). Vaccine information fact sheet for recipients and caregivers about the
Pfizer -BioNTech COVID -19 vaccine to prevent coronavirus disease 2019 (COVID -19)
for use in individuals 5 to 11 years of age. Available at:
https://www.fda.gov/media/153717/download . Accessed November 14, 2021.
20.19. FDA (2021e). Comirnaty a nd Pfizer -BioNTech COVID -19 Vaccine. Available at:
https://www.fda.gov/emergency -preparedness -and-response/coronavirus -disease -2019 -
covid -19/comirnaty -and-pfizer -biontech -covid -19-vaccine . Accessed October 30, 2021.
21.20. Ferré C, Callaghan W, Olson C, Sharma A, and Barfield W (2016). Effects of
maternal age and age -specific preterm birth rates on overall preterm birth rates —
United States, 2007 and 2014. MMWR Morb Mortal Wkly Rep;65:1181 –4.
22.21. Greenland S, Robins JM, and Pearl JS (1999). Confounding and collapsib ility in
causal inference Stat. SCI;14:29 -46.
23. Henshaw SK (1998). Unintended pregnancy in the United States. Fam Plann Perspect;
30:24 9, 46.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236372
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 42 24.22. Hernán MA and Robins JM . (201920). Casual Inference : What If , Boca Raton, FL,
Chapman & Hall/CRC.
23. Hernan MA, Brumback B, and Robins JM. (2001). Marginal Structural Models to
Estimate the Joint Causal Effect of No nrandomized Treatments. Journal of the
American Statistical Association, American Statistical Association, vol. 96, pages 440 -
448.
25.24. Hollingshead A. (1975). Four factor index of social status. Unpublished Working
Paper. Department of Sociology, Yale Univers ity, New Haven, CT. Available at:
https://sociology.yale.edu/sites/default/files/files/yjs_fall_2011.pdf#page=21 . Accessed
October 30, 2021
26. Honein MA, Paulozzi LJ, Cragan JD, and Correa A (1999). Evaluation of selected
characteristics of pregnancy drug registries. Teratology;60:356 64.
27. International Society for Pharma coepidemiology (2016) Guidelines for good
pharmacoepidemiology practice (GPP). Pharmacoepidemiol Drug Saf;25:2 10.
28. Johnson KA, Weber PA, Jones KL, Chambers CD (2001). Selection bias in Teratology
Information Service pregnancy outcome studies. Teratology;64 :7982.
29. Jones KL, Jones MC, and del Campo M (2021). Smith’s Recognizable Patterns of
Malformation, 8th Edition. Elsevier.
30. Leen Mitchell M, Martinez L, Gallegos S, Robertson J, Carey JC (2000). Mini review:
history of organized teratology information servic es in North America.
Teratology;61:314 7.
31. MMWR Morb Mortal Wkly Rep (2008). Update on overall prevalence of major birth
defects Atlanta, Georgia, 1978 2005;57:1 5.
25. Little RJA and Rubin DB. (2002). Statistical Analysis with Missing Data (2nd Ed.).
New Jersey: John Wiley & Sons, In c.
32.26. MMWR Morb Mortal Wkly Rep (2020). Birth and infant outcomes following
laboratory -confirmed SARS -CoV -2 infection in pregnancy — SET-NET, 16
Jurisdictions, March 29 –October 14; 69:1635 -40.
33. National Institutes of Health (2002). Standards for Privacy of Individually Identifiable
Health Information (45 CFR Parts 160 and 164) (Online) Available at:
https://www hhs.gov/sites/default/files/introduction.pdf. Accessed March, 2020.
34.27. Nellhaus G (1968). Head circumference from birth to eighteen years .
Pediatrics;41:106 -14.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236373
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 43 35.28. Olsen IE, Groveman SA, Lawson ML, Clark RH, and Zemel BS (2010) New
intrauterine growth curves based on United States data. Pediatrics; 125 (2):e214 -e224.
29. Pan W. (2000). A multiple imputation approach for Cox regression with interval -
censored data. Biometrics; 56: 199 -203.
36.30. Prager S, Micks E, Dalton VK (2021) Pregnancy loss (miscarriage): terminology, risk
factors and etiology. Available at: https://www.uptodate.com/ contents/pregnancy -loss-
miscarriage -clinical -presentations -diagnosis -and-initial -evaluationPregnancy loss
(miscarriage): Terminology, risk factors, and etiology. Accessed November 18, 2021.
31. Robins JM, Hernan MA, and Brumback B. (2000). Marginal Structural Models and
Causal Inference in Epidemiology. Epidemiology;11: 550 –560.
37.32. Rosenbaum PR (2002). Observational Studies (Second Edition), New York, Springer -
Verlag.
38.33. Schlaudecker EP, Munoz FM, Bardaji A, Boghossian NS, Khalil A, Mousa H, Nesin
M, Nisar MI, Pool V, Spiegel HML, Tapie MD, Kochhar S, Black S and the Brighton
Collaboration Small for Gestational Age Working Group (2017). Small for gestational
age: case definition & guidelines for data collection, analysis, and presentation of
maternal immunization safety data. Vaccine;3:6518 -6528.
39. The Center for Systems Science and Engineering at Johns Hopkins University.
Coronavirus COVID 19 Global Cases. 2021. Available at:
https://gisanddata maps.arcgis.com/apps/opsdashboard/index html#/bda7594740fd40299
423467b48e9ecf6. Accessed 17 May 2021.
40.34. Vansteelandt S and Daniel RM (2014). On regression adjustment for the propensity
score. Stat Med; 33:4053 -72.
41. World Medical Association (2013). World Medical Association Declaration of Helsinki
Ethical Principles for Medical Research Involving Human Subjects. JAMA;
310(20):2191 2194.
35. van Buuren S and Groothuis -Oudshoorn K. (2011). Mice: Multivariate imputation by
chained equations in R. Journal of Statistical Software; 45(3):1 -67.
42.36. Xu R, Hou J, Chambers CD (2018) The impact of confounder selection in propensity
scores when applied to prospective cohort studies in pregnancy. Reprod Toxicol;78:75 -
80.
37. Xu R, Luo Y, Glynn R, Johnson D, Jones KL, and Chambers C. (2014). Time -
Dependent Propensity Score for Assessin g the Effect of Vaccine Exposure on
Pregnancy Outcomes through Pregnancy Exposure Cohort Studies. International Journal
of Environmental Research and Public Health 11 (3): 3074 -3085.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236374
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 44 38. Yang S, Tsiatis AA, and Blazing M. (2018). Modeling survival distribution as a
function of time to treatment discontinuation: A dynamic treatment regime approach.
Biometrics. Sep;74(3):900 -909.
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236375
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 45 14. LIST OF TABLES
Table 1. Sample Size and Power for a Specified Effect Size ................................ .12
Table 2. Variables Related to Defining the Exposure, Comparator,
Outcomes, Exclusions, and Date/Time ................................ ..................... 17
Table 3. Demographic and Clinical Variables ................................ ........................ 23
Table 4. Denominators for Outcomes by Exposure/Comparator Cohort ............... 28
15. LIST OF FIGURES
None.
ANNEX 1. LIST OF STAND -ALONE DOCUMENTS
None.
ANNEX 2. ADDITIONAL INFORMATION
Table A -1. List of known human teratogens
Exposure Notes
Angiotensin -converting enzyme
inhibitors/ Angiotensin II receptor
blockers Exposure in the 2nd or 3rd trimester
Acitretin Any exposure within 2 years of LMP.
Alcohol Use, Heavy Must average 14 or more standard drinks over
four or more weeks post -conception
Aminopterin Exposure occurring on or after LMP
Antiseizure / Anticonvulsant
Medications Exposure occurring on or after LMP
Antineoplastics, Other Exposure occurring on or after LMP
Cocaine Exposure occurring on or after LMP
Cytomegalovirus (CMV) Primary infection occurring on or after LMP
Type I and Type II Diabetes Any diagnosis
Etretinate Any exposure within 10 years of LMP.
Fever, High 102 degrees or higher for 24 hours or longer on
or after LMP
Fluconazole, Systemic >7 days total (consecutive or non -consecutive) on
or after LMP
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236376
Pfizer -BioNTech COVID -19 Vaccine
C4591022 NON -INTERVENTIONAL STUDY STATISTICAL ANALYSIS PLAN
Version 12.0, 30 November 2021 05 May 2022
PFIZER CONFIDENTIAL
Page 46 Table A -1. List of known human teratogens
Exposure Notes
Isotretinoin Exposure occurring on or after LMP
Lenalidomide Exposure occurring on or after LMP
Lithium Exposure occurring on or after LMP
Methimazole Exposure occurring on or after LMP
Methotrexate Exposure occurring on or after LMP
Propylthiouracil (PTU) Exposure occurring on or after LMP
Radiation, High Dose > 5 rads to the uterus on or after LMP
Rubella Exposure occurring on or after LMP
Thalidomide Exposure occurring on or after LMP
Toxoplasmosis Primary infection occurring on or after LMP
Varicella Primary infection occurring on or after LMP
Vitamin A, High Dose >50,000 IU per day on or after LMP
Warfarin (Coumadin, Jantoven)
derivatives Exposure occurring on or after LMP
Zika, Confirmed Positive test result
090177e19a40b34a\Final\Final On: 19-May-2022 13:59 (GMT)
FDA-CBER-2022-5812-0236377
FDA-CBER-2022-5812-0236378