search

CoVPN / correlates_reporting_usgcove_archive

Reproducible reporting workflows for the immune correlates statistical analyses of the Moderna and Janssen COVID-19 vaccine efficacy trials by the USG/COVE Response Biostatistics Team [Archived 16 October 2021]
GNU General Public License v3.0
6 stars 15 forks source link

`TwophasesampInd` definition #297

Closed benkeser closed 3 years ago

benkeser commented 3 years ago

@youyifong, is this line correct or should it be EventIndPrimaryD57?

benkeser commented 3 years ago

Another question: I see this note in data_raw/README.md:

But I don't think that's true right? Because someone could have an event >=7 days post day 29 but less than >=7 days post day 57. What am I missing?

youyifong commented 3 years ago

You are right. This readme.md is unfortunately out of date. I have attached the most recent version I have. Peter may have a newer one.

On Thu, May 6, 2021 at 8:16 AM David Benkeser @.***> wrote:

Another question: I see this note in data_raw/README.md https://github.com/CoVPN/correlates_reporting/blob/master/data_raw/README.md :

  • EventIndPrimaryD29==1 implies EventIndPrimaryD57==1

But I don't think that's true right? Because someone could have an event

=7 days post day 29 but less than >=7 days post day 57. What am I missing?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/CoVPN/correlates_reporting/issues/297#issuecomment-833605222, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHBH5LYX2WAMLUCGE7FHSNLTMKXDNANCNFSM44HGCNIQ .

README File: Mock COVID-19 Vaccine Efficacy Trial data sets COVID_VEtrial_practicedata_primarystage1.csv COVID_VEtrial_practicedata_longerterm.csv

     [In Q1 2021 only need to consider COVID_VEtrial_practicedata_primarystage1.csv]

Author: Peter Gilbert, September 16, 2020 Updated October 9, 2020 Updated October 25, 2020 Updated November 12, 2020 Updated November 28, 2020 Updated December 17, 2020 (mainly added variables related to studying Day 29 markers) Updated January 3, 2021 (added plotting/tabulating terminology at the bottom and clarified variables) Updated January 21, 2021 (Added binding Ab to Nucleocapsid protein (N) for immunogenicity reporting, changed wildtype live-virus neutralization variables to MN50 (Battelle assay), and edits for enhancing clarity) Updated April 15, 2021 (updated LLOD, LLOQ, ULOQ values, including conversion of bAb readouts to the International Units/ml scale) Updated April 23, 2021 (Changed documentation of Perprotocol variable) Updated April 30, 2021 (Added the variables EarlyinfectionD29, EarlyinfectionD57, NumberdaysD1toD29, NumberdaysD1toD57; URMforsubcohortsampling (derived from elemental race and ethnicity variables); updated LLOD, LLOQ, ULOQ values for bAb)

This README file describes the variables in the mock data sets COVID_VEtrial_practicedata_primarystage1.csv and COVID_VEtrial_practicedata_longerterm.csv

The first data set represents an expected data set that would be available for the first Stage 1 correlates analysis of a Phase 3 trial, which takes place at some point after the primary analysis of vaccine efficacy that took place after at least 150 COVID primary endpoint cases. The second data set is the same as the first data set, except updated to include approximately 6 months of additional follow-up, with many more endpoint cases for analysis.

The data sets represent a mock example Phase 3 data set for a CoVPN/OWS Phase 3 COVID-19 vaccine efficacy trial, that would be used for Day 57 antibody marker correlates of risk and correlates of protection analyses. The primary endpoint is virologically confirmed COVID-19 disease (e.g., Krause et al., Lancet, 2020; Mehrotra et al., 2020, Ann Int Med).
The data sets fit the Moderna trial design.

The trial randomizes individuals to vaccine or placebo (administered at Day 1 and Day 29) and follows participants for 25 months for occurrence of the COVID-19 primary endpoint and for secondary endpoints. The randomization allocates 1:1 to vaccine:placebo. The study design was event driven, with the primary analysis triggered at 150 COVID-19 primary endpoints combined over the vaccine and placebo arms, although the correlates analysis takes place after that with a greater number of cases. The simulated data set uses an estimate of vaccine efficacy of 80% in the primary analysis cohort, assuming that the primary analysis did not occur earlier based on reaching a monitoring boundary at an interim analysis.

The data set considers the standard CoVPN/OWS two-phase sampling case-cohort design for sampling antibody markers at Day 1 (baseline/enrollment), at Day 29 (second vaccination), and at Day 57 (approximate peak antibody time point after the second vaccination at Day 29). This two-phase sampling case-cohort design is described in a separate document (to include on the Github page) and summarized in the immune correlates SAP.

The case-cohort design measures antibody markers at (Day 1, Day 29, Day 57) using three immunoassays, which measure (1) binding antibody to the vaccine-insert Spike protein and to the receptor binding domain (RBD) protein (two readouts), (2) pseudovirus neutralizing antibody titer to the vaccine-insert spike protein, and (3) live virus neutralizing antibody titer to the vaccine-insert spike protein.

The first level of immunological biomarker peak antibody correlates analyses assesses "Day 57 Correlates of Risk" in SARS-CoV-2 baseline seronegative vaccine recipients, where a correlate of risk is an antibody marker measured at Day 57 that associates with subsequent occurence of the COVID-19 endpoint, adjusting for baseline covariates. While correlates of risk analyses are restricted to baseline seronegative vaccine recipients, the entire data set is simulated, so that graphical descriptive immunogenicity analyses can also be conducted, which include data from both the vaccine and placebo groups and from baseline negative and baseline positive participants. In addition, additional analyses are done in the pursuit of a correlate of protection (statistical frameworks for correlates of protection are summarized in the immune correlates SAP).

Parallel analyses are done for assessing Day 29 correlates of risk and protection. In addition, the multivariable correlate of risk objective considers Day 29 and Day 57 markers in the same regression models.

Variables in the Data Set COVID_VEtrial_practicedata_primarystage1.csv:

Trt = Randomized treatment assignment (1=vaccine, 0=placebo) EthnicityHispanic = Indicator ethnicity = Hispanic (0 = Non-Hispanic) EthnicityNotreported = Inicator ethnicity = Not reported (0 = Non-Hispanic) EthnicityUnknown = Indicator ethnicity = Unknown (0 = Non-Hispanic) Black = Indicator race = Black (0 = White) Asian = Indicator race = Asian (0 = White) NatAmer = Indicator race = American Indian or Alaska Native (0 = White) PacIsl = Indicator race = Native Hawaiian or Other Pacific Islander (0 = White) Multiracial = Indicator race = Multiracial (0 = White) Other = Indicator race = Other (0 = White) Notreported = Indicator race = Not reported (0 = White) Unknown = Indicator race = unknown (0 = White) URMforsubcohortsampling = Indicator of under-represented minority (1=Yes, 0=No, NA) This variable matches the sampling implemented by Moderna. Minority is defined as: Blacks or African Americans, Hispanics or Latinos, American Indians or Alaska Natives, NativeHawaiians, and other Pacific Islanders. Non-Minority includes all the others whose race (i.e. Asian, Multiracial, Other) or ethnicity is not unknown, unreported or missing. Therefore Unknown and Not reported have NA for this variable. RiskInd = Baseline covariate high risk/at-risk pre-existing condition (1=yes, 0=no) Sex = Sex assigned at birth (1=female, 0=male) Age = Age at enrollment in years, between 18 and 85. Note that the randomization strata are 18-64 and 65+. BMI = BMI at enrollment (kg/m^2) NumberdaysD1toD29 = number of days between D1 visit and Day 29 visit NumberdaysD1toD57 = number of days between D1 visit and Day 57 visit Bserostatus = Indicator of baseline SARS-CoV-2 seropositive (1=yes, 0=no) Fullvaccine = Indicator of receipt of both vaccinations. Perprotocol = Indicator of qualifying per-protocol (received both vaccinations without specified protocol violations, and not diagnosed with the COVID endpoint by the Day 29 / Dose 2 visit). Importantly, this indicator may differ from the Per-protocol indicator in the clinical data base for the primary analysis of vaccine efficacy, because that indicator may be zero for COVID cases with event time prior to 14 days post dose 2. Perprotocol is a cohort-qualification indicator that is used for defining inclusion of participants for immunogenicity and immune correlates analyses. EventTimePrimaryD1 = Minimum of of the time from Day 1 (first dose) until the COVID-19 endpoint or right-censoring (in days). For ITT analysis. EventIndPrimaryD1 = Indicator that the ITT failure time is <= the right-censoring time. EventTimePrimaryD29 = Minimum of of the time from Day 29 (antibody marker measurement) until the COVID-19 endpoint or right-censoring (in days). Note that Day 29 is the time origin for studying Day 57 antibody markers as correlates. EventIndPrimaryD29 = Indicator that the failure time is <= the right-censoring time. Note that COVID-19 endpoints are only counted starting 7 days post Day 29 visit, because endpoints occurring over the first 6 days may have already been infected with SARS-CoV-2 before endpoint occurrence. This means that all failure events included in the analysis have failure time >= 7 days. Individuals with failure time 1 to 6 days are excluded from the analysis. EventTimePrimaryD57 = Minimum of of the time from Day 57 (antibody marker measurement) until the COVID-19 endpoint or right-censoring (in days). Note that Day 57 is the time origin for studying Day 57 antibody markers as correlates. EventIndPrimaryD57 = Indicator that the failure time is <= the right-censoring time. Note that COVID-19 endpoints are only counted starting 7 days post Day 57 visit, because endpoints occurring over the first 6 days may have already been infected with SARS-CoV-2 before endpoint occurrence. This means that all failure events included in the analysis have failure time >= 7 days. Individuals with failure time 1 to 6 days are excluded from the analysis. BbindSpike = Day 1 (enrollment) value of log10 IgG binding antibody concentration to Spike protein, which is a continuous variable (scale log10 IU/ml).
BbindRBD = Day 1 (enrollment) value of IgG binding antibody readout to RBD, in IU/ml. BbindN = Day 1 (enrollment) value of IgG binding antibody readout to N protein, in IU/ml. Bpseudoneutid50 = Day 1 value of the Duke pseudo-neutralizing antibody readout, reported as log10 estimated serum inhibitory dilution 50% titer (ID50), which is the reciprocal of the dilution at which RLU (relative luminescence units) are reduced by either 50% (ID50) compared to virus control wells after subtraction of background RLUs in cell control wells.
Bpseudoneutid80 = Day 1 value of the Duke pseudo-neutralizing antibody readout, reported as log10 estimated serum inhibitory dilution 80% titer (ID80), which is the reciprocal of the dilution at which RLU (relative luminescence units) are reduced by either 80% (ID80) compared to virus control wells after subtraction of background RLUs in cell control wells.
BliveneutMN50 = Day 1 value of Battelle assay wild type live virus-neutralizing antibody readout, reported as log10 MN50. Day29bindSpike = Day 29 value of the same marker as BbindSpike Day29bindRBD = Day 29 value of the same marker as BbindRBD Day29bindN = Day 29 value of the same marker as BbindN Day29pseudoneutid50 = Day 29 value of the same marker as Bpseudoneutid50 Day29pseudoneutid80 = Day 29 value of the same marker as Bpseudoneutid80 Day29liveneutmn50 = Day 29 value of the same marker as Bliveneutmn50 Day57bindSpike = Day 57 value of the same marker as BbindSpike Day57bindRBD = Day 57 value of the same marker as BbindRBD Day57bindN = Day 57 value of the same marker as BbindN Day57pseudoneutid50 = Day 57 value of the same marker as Bpseudoneutid50 Day57pseudoneutid80 = Day 57 value of the same marker as Bpseudoneutid80 Day57liveneutmn50 = Day 57 value of the same marker as Bliveneutmn50 SubcohortInd = Indicator that a participant is sampled into the random subcohort for measurement of Day 1, 29, 57 antibody markers (stratified Benoulli random sampling) EarlyinfectionD29 = Indicator a participant has SARS-CoV-2 infection < 7 days post Day 29 visit (0 otherwise). EarlyinfectionD57 = Indicator a participant has SARS-CoV-2 infection < 7 days post Day 57 visit (0 otherwise).

Note that infection may be detected by RNA PCR/NAAT or by seroconversion.

The remaining variables with "CPV" in the variable name are only used for correlate of VE CoP analysis:

BbindSpikeCPV = Day 1 value of log10 IgG binding antibody readout to Spike protein in non-case placebo recipients undergoing closeout placebo vaccination. NA if the value is not measured. BbindRBDCPV = Day 1 value of log10 IgG binding antibody readout to RBD in non-case placebo recipients undergoing closeout placebo vaccination. NA if the value is not measured. Bpseudoneutid50CPV = Day 1 value of pseudo neutralizing antibody log10 ID50 readout in non-case placebo recipients undergoing closeout placebo vaccination. NA if the value is not measured. Bpseudoneutid80CPV = Day 1 value of pseudo neutralizing antibody log10 ID80 readout in non-case placebo recipients undergoing closeout placebo vaccination. NA if the value is not measured. Bliveneutmn50CPV = Day 1 value of wild type live virus neutralizing antibody log10 MN50 readout in non-case placebo recipients undergoing closeout placebo vaccination. NA if the value is not measured. Day29SbindSpikeCPV = Same as BbindSpikeCPV, at Day 29 (thus is a post-vaccination response). Day29SbindRBDCPV = Same as BbindRBDCPV, at Day 29 Day29Spseudoneutid50CPV = Same as Bpseudoneutid50CPV, at Day 29 Day29Spseudoneutid80CPV = Same as Bpseudoneutid80CPV, at Day 29 Day29Sliveneutmn50CPV = Same as Bliveneutid50CPV, at Day 29 Day57SbindSpikeCPV = Same as BbindSpikeCPV, at Day 57 (thus is a post-vaccination response). Day57SbindRBDCPV = Same as BbindRBDCPV, at Day 57 Day57Spseudoneutid50CPV = Same as Bpseudoneutid50CPV, at Day 57 Day57Spseudoneutid80CPV = Same as Bpseudoneutid80CPV, at Day 57 Day57Sliveneutmn50CPV = Same as Bliveneutmn50CPV, at Day 57 CPVsampInd = Indicator that a participant is sampled into the closeout placebo vaccination cohort and has (Day 1, Day 29, Day 57) antibody marker measurements. NA if a vaccine recipient or a placebo recipient failure event/case. Relevant for correlate of VE methods. Every placebo recipient with CPVsampInd = 1 has data on all of the Day 1, 29, 57 antibody marker variables.

#################################################################################

Data analysis notes

EventIndPrimaryD57==1 implies EventIndPrimaryD29==1 implies EventIndPrimaryD1==1

Based on the data set, a new variable TwophasesampIndprimary is defined for the processed data set, which is the indicator that a participant has antibody marker data measured, where all participants with TwophasesampIndprimary==1 have complete data for Day 1, 29, 57 markers, and all immunogenicity report analyses restrict to ptids with TwophasesampIndprimary==1.
In addition, all CoR/CoP analyses of Day 57 markers only, and of Day 29 and Day 57 markers together, restrict to ptids with TwophasesampIndprimary==1. All CoR/CoP analyses of Day 29 markers restrict to ptids with TwophasesampIndprimary.2==1, where all participants with TwophasesampIndprimary.2==1 have complete data for Day 1, 29 markers. The reason a new variable is used is because some intercurrent COVID cases (before 6 days post Day 57 visit) may miss the Day 57 visit due to COVID.

Two other derived variables are helpful to define ptids to include in analyses EarlyendpointD57 = Indicator a participant has EarlyinfectionD57==1 or has a COVID endpoint < 7 days post Day 57 visit (0 otherwise) [used to exclude ptids from the immunogenicity analyses and to exclude non-cases from CoR/CoP analyses] This variable is derived as EarlyendpointD57 <- ifelse(EarlyinfectionD57==1 | (EventIndPrimaryD1==1 & EventTimePrimaryD1 < NumberdaysD1toD57 + 7),1,0)

For parallel structure, we also include: EarlyendpointD29 = Indicator a participant has EarlyinfectionD29==1 or has a COVID endpoint < 7 days post Day 29 visit (0 otherwise)
EarlyendpointD29 <- ifelse(EarlyinfectionD29==1 | (EventIndPrimaryD1==1 & EventTimePrimaryD1 < NumberdaysD1toD29 + 7),1,0)

CoR/CoP analyses of Day 29 markers only:
Phase 1 ptids: Bserostatus==0 & EarlyendpointD29==0 & Perprotocol==1 & EventTimePrimaryD29 >= 7
Phase 2 ptids: TwophasesampIndprimary.2==1 Failure time variables: (EventTimePrimaryD29, EventIndPrimaryD29) and ignore (EventTimePrimaryD57, EventIndPrimaryD57)

CoR/CoP analyses of Day 57 markers only:
Phase 1 ptids: Bserostatus==0 & EarlyendpointD57==0 & Perprotocol==1 & EventTimePrimaryD57 >= 7 Phase 2 ptids: TwophasesampIndprimary==1 Failure time variables: (EventTimePrimaryD57, EventIndPrimaryD57) and ignore (EventTimePrimaryD29, EventIndPrimaryD29)

CoR/CoP analyses of Day 29 and Day 57 markers together:
Phase 1 ptids: Bserostatus==0 & EarlyendpointD57==0 & Perprotocol==1 & EventTimePrimaryD57 >= 7 Phase 2 ptids: TwophasesampIndprimary==1 Failure time variables: (EventTimePrimaryD57, EventIndPrimaryD57) and ignore (EventTimePrimaryD29, EventIndPrimaryD29)

Immunogenicity report analyses include ptids with Perprotocol==1 & SubcohortInd==1 & TwophasesampIndprimary==1 & EarlyendpointD57==0

Intercurrent Cases are defined as cases with event time between 7 days post Day 29 visit and 6 days post Day 57 visit. Intercurrent cases have included ptids defined by Perprotocol==1 & Bserostatus==0 & EarlyendpointD29==0 & TwophasesampIndprimary.2==1 & EventIndPrimaryD29==1 & EventTimePrimaryD29 >= 7 & EventIndPrimaryD57==0

Per-protocol non-cases are defined as participants who never register a COVID primary endpoint and have no evidence of infection, with ptids for IPW-complete case analyses defined by Perprotocol==1 & Bserostatus==0 & EarlyendpointD57==0 & TwophasesampIndprimary==1 & EventIndPrimaryD1==0

################### Descriptive plots in CoR and CoP reports (e.g. violin scatterplots) include "Intercurrent Cases", "Primary Cases", "Non-Cases", side by side.

Intercurrent Cases ptids to include in plots: Perprotocol==1 & Bserostatus==0 & EarlyendpointD29==0 & TwophasesampIndprimary.2==1 & EventIndPrimaryD29==1 & EventTimePrimaryD29 >= 7 & EventIndPrimaryD57==0

Primary Cases ptids to include in plots: Perprotocol==1 & Bserostatus==0 & EarlyendpointD57==0 & TwophasesampIndprimary==1 & EventIndPrimaryD57==1 & EventTimePrimaryD57 >= 7

Non-cases ptids to include in plots: Perprotocol==1 & Bserostatus==0 & EarlyendpointD57==0 & TwophasesampIndprimary==1 & EventIndPrimaryD1==0

################# Subgroup analyses for both immunogenicity reporting and CoR/CoP reporting by race ethnicity are based on the URMforsubcohortsampling variable derived in the processed data set. IPW weights are also computed using this variable, because the subchort sampling design used this variable.

URMforsubcohortsampling = Indicator of under-represented minority (URM) (1, 0, or NA). This variable matches the subcohort sampling implemented by Moderna. Minority is defined as: Blacks or African Americans, Hispanics or Latinos, American Indians or Alaska Natives, NativeHawaiians, and other Pacific Islanders. Non-Minority includes all the others whose race (i.e. Asian, Multiracial, Other) or ethnicity is not unknown, unreported or missing. Therefore Unknown and Not reported have NA for this sampling stratum variable.

URMforsubcohortsampling <- rep(NA,length(A)) URMforsubcohortsampling[Black==1] <- 1 URMforsubcohortsampling[EthnicityHispanic==1] <- 1 URMforsubcohortsampling[NatAmer==1] <- 1 URMforsubcohortsampling[PacIsl==1] <- 1 URMforsubcohortsampling[Asian==1 & EthnicityHispanic==0] <- 0 URMforsubcohortsampling[Multiracial==1 & EthnicityHispanic==0] <- 0 URMforsubcohortsampling[Other==1 & EthnicityHispanic==0] <- 0

Add observed White Non Hispanic:

URMforsubcohortsampling[EthnicityHispanic==0 & Black==0 & Asian==0 & NatAmer==0 & PacIsl==0 & Multiracial==0 & Other==0 & Notreported==0 & Unknown==0] <- 0

Covariate-adjustment in CoR/CoP analyses adjusts for a WhiteNonHispanic variable that is defined differently from URMforsubcohortsampling. WhiteNonHispanic==1 if observed White and observed Not Hispanic or Latino and is referred to as "White Non-Hispanic". WhiteNonHispanic==0 if observed non-White and/or observed Hispanic. In addition, for ptids with not reported or unknown on White Non-Hispanic status, the assigned value is 1. Thus this variable has no NAs, which is convenient for use in covariate-adjustment.

##########################################################################

In analyzing the marker data, customs in handling the lower limit of detection (LLOD) need to

be handled in the analysis code.

Analyses of baseline/Day 1 markers, and of Day 29 and Day 57 markers, assign all

values < LLOD to the value LLOD/2. This is an initial step in the data processing to create

an analysis ready data set.

Immunogenicity analyses use all actual values above the ULOQ.

CoR/CoP analyses assign values above the ULOQ to the ULOQ.

The LLODs, ULODs, LLOQs and ULOQs are as follows (on the natural/antilog10 scale):

Marker              LLOD    ULOD        LLOQ    ULOQ

bindSpike (IU/ml scale) 0.3076 172,226.2 1.7968 10,155.95

bindRBD (IU/ml scale) 0.9297 520,506.0 5.4302 30,693.537

bindN (IU/ml scale) 0.0820 45,927.0 0.4791 2708.253

pseudoneutid50 10 - 18.5 4404

pseudoneutid80 10 - 14.3 1295

livevirusneutmn50 62.16 - 117.35 18,976.19

The VRC has a report on the Conversion of SARS-CoV-2 Binding Assay Results from Arbitrary Units to WHO International Units (draft March 30, 2021), which establishes conversion factors for the MSD AU/ml units to WHO International Units/ml.
For the three binding antibody variables CoV-2 Spike IgG, CoV-2 N IgG, and CoV-2 RBD IgG, these conversion factors are 0.0090, 0.0024, and 0.0272, respectively. These conversion factors are applied, such that all binding Ab readouts are reported in WHO IU/ml. These conversion factors from AU/ml to IU/ml are applied to yield the LLOD, LLOQ, and ULOQ on the WHO IU/ml scale that are provided in the table above.

################################################################################# Note 1/21/21: The live virus neutralization data are available later than the pseudo virus neutralization data. Therefore, for the initial reports, the following 4 markers are focused on for immunogenicity figures bindSpike, bindRBD, pseudoneutid50, pseudoneutid80 and the following 5 markers are focused on for immunogenicity tables and figures bindSpike, bindRBD, bindN, pseudoneutid50, pseudoneutid80. The first CoR/CoP reports focus on the 4 markers bindSpike, bindRBD, pseudoneutid50, pseudoneutid80 #################################################################################

Variables in the Data Set COVID_VEtrial_practicedata_longerterm.csv:

Identical to the first data set, except for the following modifications:

EventTimePrimaryD29 is replaced with EventTimeLongtermD29, based on ~6 months additional follow-up
EventIndPrimaryD29 is replaced with EventIndLongtermD29, based on ~6 months additional follow-up

EventTimePrimaryD57 is replaced with EventTimeLongtermD57, based on ~6 months additional follow-up
EventIndPrimaryD57 is replaced with EventIndLongtermD57, based on ~6 months additional follow-up

All of the antibody biomarker variables are the same, except the variables for the longer term data set with TwophasesampIndprimary==NA converted to TwophasesampIndLongterm==1 almost always have a value for the variables (given that the antibody markers are measured in all additional cases).

###################################################################################

Notes on the two-phase sampling case-cohort design (Prentice, 1986, Biometrika; Breslow et al., 2009, AJE):

There are two kinds of variables: "phase one variables" measured in everyone and "phase two variables" only measured in the case-cohort sample for which antibody data are measured. For Moderna, this consists of the random subcohort with 24 baseline strata defined by Trt x URMforsubcohortsampling (Yes,No) x [age >= 65, age < 65 & HighRiskInd==1, age < 65 & HighRiskInd==0] x Bserostatus.

All variables are phase one variables except the 18 antibody marker variables are phase two variables (BbindSpike, BbindRBD, BbindN, Bpseudoneutid50, Bpseudoneutid80, Bliveneutmn50, Day29bindSpike, Day29bindRBD, Day29bindN, Day29pseudoneutid50, Day29pseudoneutid80, Day29liveneutmn50, Day57bindSpike, Day57bindRBD, Day57N, Day57pseudoneutid50, Day57pseudoneutid80, Day57liveneutmn50).

A variable TwophasesampInd is derived from the data set, which is the indicator that a ppt has Day 1, 29, 57 marker data and hence is included in inverse probability weighting (IPW) complete case data analysis. The two-phase sampling probabilities of trial participants,

P(TwophasesampInd=1|Trt,WhiteNonHispanic Indicator,HighRiskInd,Age Category,Bserostatus,EventIndPrimaryD29),

can be consistently estimated based on empirical sampling frequencies. Note that the weights depend on case/non-case COVID outcome status. Note that the URMforsubcohortsampling Indicator is derived from the other ethnicity and race variables.

A second variable TwophasesampInd.2 is also derived, which is only used as an inclusion indicator for CoR/CoP analyses that only include Day 29 markers.

For the longer term follow-up data set, another variable TwophaseLongtermsampInd is derived from the data set, which is the indicator that a ppt has Day 1, 29, 57 marker data and hence is included in IPW complete case data analysis (the additional cases are added). Again, the two-phase sampling probabilities of trial participants,

P(TwophaseLongtermsampInd=1|Trt,URMforsubcohortsampling Indicator,HighRiskInd,Age Category,Bserostatus,EventIndLongtermD29),

can be consistently estimated based on empirical sampling frequencies.

Notes 11/12/2020: The Moderna data set will be analyzed first. Therefore, the results should be reported using two phase sampling probabilities based on the Moderna design. Sampling was done based on the strata (Age >= 65, Age < 65 and at risk, Age < 65 and not at risk) x (URMforsubcohortsampling Yes, No) x (baseline negative, baseline positive) x (vaccine, placebo). For reporting of subgroups, the following comparisons are made, where the terms indicate the labels that should be used.

  1. Age < 65 vs. Age >= 65
  2. At risk vs. Not at risk
  3. Age < 65 at risk vs. Age < 65 not at risk
  4. Age >= 65 at risk vs. Age >= 65 not at risk
  5. Male vs. Female
  6. White Non-Hispanic vs. Comm. of color
  7. Hispanic or Latino vs. Not Hispanic or Latino

Notes regarding Ethnicity and Race variables:

There are 4 total levels of EthnicityHispanic:

  1. Hispanic or Latino
  2. Not Hispanic or Latino
  3. Not reported
  4. Unknown

Tabular results report for the three subgroups 1., 2., and (3. and 4.) combined (as done in the Baden et al. 2020 NEJM paper). The three subgroups reported on are defined as follows, with these labels:

  1. Hispanic or Latino: EthnicityHispanic==1
  2. Not Hispanic or Latino: EthnicityHispanic==0 & EthnicityNotreported==0 & EthnicityUnknown==0
  3. Not reported and unknown: (EthnicityNotreported==1 | EthnicityUnknown==1)

All subgroups labeled White are defined by the Race variable indicating White and the EthnicityHispanic variable indicating Not Hispanic or Latino.
This is coded as the indicator of Race being White and Ethnicity being Not Hispanic as: WhiteNonHispanic <- ifelse(EthnicityHispanic==0 & EthnicityNotreported==0 & EthnicityUnknown==0 & Black==0 & Asian==0 & NatAmer==0 & PacIsl==0 & Multiracial==0 & Notreported==0 & Other==0 & Unknown==0,1,0)
WhiteNonHispanic[(EthnicityNotreported==0 & EthnicityUnknown==0) & Unknown==1] <- 1

For labeling, the Communities of Color subgroup (label 'Comm. of Color'), defined by URMforsubcohortsampling==1, uses label 'Comm. of Color' or 'Communities of Color'. The White Non-Hispanic subgroup is defined by URMforsubcohortsampling==0 and is labeled by 'White Non-Hispanic'.

There are 9 total levels of Race:

  1. White
  2. Black
  3. Asian
  4. NatAmer
  5. PacIsl
  6. Multiracial
  7. Other
  8. Notreported
  9. Unknown

Tabular results report for the 8 subgroups 1., 2., ..., 7. and (8. and 9.) combined (as done in the Baden et al. 2020 NEJM paper). The eight subgroups reported on are defined as follows, with these labels:

  1. White Non-Hispanic (so White is actually defined by WhiteNonHispanic==1)
  2. Black or African American
  3. Asian
  4. American Indian or Alaska Native
  5. Native Hawaiian or Other Pacific Islander
  6. Multiracial
  7. Other
  8. Not reported and unknown

Short labels if the above labels are too long:

  1. White Non-Hispanic
  2. Black
  3. Asian
  4. NatAmer
  5. PacIsl
  6. Multiracial
  7. Other
  8. NA/unknown

Antibody marker labeling for all figures and tables:

Axis Labels Title Labels Anti Spike IgG (IU/ml) Binding Antibody to Spike: Day 1 Anti RBD IgG (IU/ml) Binding Antibody to RBD: Day 1 Anti N IgG (IU/ml) Binding Antibody to N: Day 1 Pseudovirus-nAb ID50 Pseudovirus Neutralization ID50: Day 1 Pseudovirus-nAb ID80 Pseudovirus Neutralization ID80: Day 1 Live virus-nAb MN50 Live virus Neutralization MN50: Day 1

Anti Spike IgG (IU/ml) Binding Antibody to Spike: Day 29 Anti RBD IgG (IU/ml) Binding Antibody to RBD: Day 29 Anti N IgG (IU/ml) Binding Antibody to N: Day 29 Pseudovirus-nAb ID50 Pseudovirus Neutralization ID50: Day 29 Pseudovirus-nAb ID80 Pseudovirus Neutralization ID80: Day 29 Live virus-nAb MN50 Live virus Neutralization MN50: Day 29

Anti Spike IgG (IU/ml) Binding Antibody to Spike: Day 57 Anti RBD IgG (IU/ml) Binding Antibody to RBD: Day 57 Anti N IgG (IU/ml) Binding Antibody to N: Day 57 Pseudovirus-nAb ID50 Pseudovirus Neutralization ID50: Day 57 Pseudovirus-nAb ID80 Pseudovirus Neutralization ID80: Day 57 Live virus-nAb MN50 Live virus Neutralization MN50: Day 57

Anti Spike IgG (IU/ml) Binding Ab to Spike: D29 fold-rise over D1 Anti RBD IgG (IU/ml) Binding Ab to RBD: D29 fold-rise over D1 Anti N IgG (IU/ml) Binding Ab to N: D29 fold-rise over D1 Pseudovirus-nAb ID50 Pseudovirus nAb ID50: D29 fold-rise over D1 Pseudovirus-nAb ID80 Pseudovirus nAb ID80: D29 fold-rise over D1 Live virus-nAb MN50 Pseudovirus nAb MN50: D29 fold-rise over D1

Anti Spike IgG (IU/ml) Binding Ab to Spike: D57 fold-rise over D1 Anti RBD IgG (IU/ml) Binding Ab to RBD: D57 fold-rise over D1 Anti N IgG (IU/ml) Binding Ab to N: D57 fold-rise over D1 Pseudovirus-nAb ID50 Pseudovirus nAb ID50: D57 fold-rise over D1 Pseudovirus-nAb ID80 Pseudovirus nAb ID80: D57 fold-rise over D1 Live virus-nAb MN50 Pseudovirus nAb MN50: D57 fold-rise over D1

Risk of symptomatic COVID Intercurrent cases PP Cases PP Non-cases Sex assigned at birth by race and ethnic group by dichotomous classification of race and ethnic group

Pseudovirus-nAb ID50
Pseudovirus-nAb ID80
Wild type live-virus nAb MN50 Dayx Pseudovirus-nAb ID50 Dayx Pseudovirus-nAb ID80 Dayx Wild type live-virus nAb MN50 where x=1, 29, or 57.

Pseudovirus-nAb ID50: D29 fold-rise over D1 Pseudovirus-nAb ID80: D29 fold-rise over D1 Wild type live-virus nAb MN50: D29 fold-rise over D1 Pseudovirus-nAb ID50: D57 fold-rise over D1 Pseudovirus-nAb ID80: D57 fold-rise over D1 Wild type live-virus nAb MN50: D57 fold-rise over D1

Short-hand notation when needed: Pseudovirus-nAb shortened to PsV-nAb Wild type live-virus nAb shortened to WT LV-nAb

youyifong commented 3 years ago

This was intentional. My understanding is that this would cover intercurrent cases.

On Thu, May 6, 2021 at 7:41 AM David Benkeser @.***> wrote:

@youyifong https://github.com/youyifong, is this line https://github.com/CoVPN/correlates_reporting/blob/623b03084c73a1080f8b33c1e3e470c2e280cd96/data_clean/make_dat_proc.R#L51 correct or should it be EventIndPrimaryD57?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/CoVPN/correlates_reporting/issues/297, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHBH5L3ZQ5O4LAJPXUDF3HLTMKTC3ANCNFSM44HGCNIQ .

benkeser commented 3 years ago

This was intentional. My understanding is that this would cover intercurrent cases.

Thanks. So it must be true that all Day 57 analysis scripts need to explicitly exclude intercurrent cases rather than simply subsetting on TwophasesampInd == 1

youyifong commented 3 years ago

TwophasesampInd is refined later in the script. For D57, I am using is.na(wt) to get ph1 ptids and is.na(wt) & TwophasesampInd to find ph2 ptid. For D29, it is wt.2 and TwophasesampInd.2. I should probably have chosen .D57 and .D29 to distinguish the two.

On Thu, May 6, 2021 at 8:26 AM David Benkeser @.***> wrote:

This was intentional. My understanding is that this would cover intercurrent cases.

Thanks. So it must be true that all Day 57 analysis scripts need to explicitly exclude intercurrent cases rather than simply subsetting on TwophasesampInd == 1

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/CoVPN/correlates_reporting/issues/297#issuecomment-833613165, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHBH5L5TLCUUXVQSD5JHBTDTMKYL7ANCNFSM44HGCNIQ .

benkeser commented 3 years ago

Ok thanks