FDA Adverse Event Reporting System (FAERS)

Github Actions Badge

The post-marketing safety surveillance program for drug and therapeutic biological products.

  • Multiple tables linked by primaryid including demographics, outcomes, drug start and end dates.

  • Voluntary reports from practitioners and patients, not representative, no verification of causality.

  • Published quarterly since 2004, file structure revisions at 2012Q4 and 2014Q3.

  • Maintained by the United States Food and Drug Administration (FDA).


Please skim before you begin:

  1. ASC_NTS.DOC included in each quarterly zipped file, especially the Entity Relationship Diagram

  2. Questions and Answers on FDA’s Adverse Event Reporting System (FAERS)

  3. This human-composed haiku or a bouquet of artificial intelligence-generated limericks

# side effect guestbook
# violet you're turning violet
# vi'lent dose response

Download, Import, Preparation

Download the quarterly file:

library(httr)

tf <- tempfile()

this_url <- "https://fis.fda.gov/content/Exports/faers_ascii_2023q1.zip"

GET( this_url , write_disk( tf ) , progress() )

unzipped_files <- unzip( tf , exdir = tempdir() )

Define a function to import each text file:

read_faers <-
    function( this_fn ){
        read.table( this_fn , sep = "$" , header = TRUE , comment.char = "" , quote = "" )
    }

Import multiple tables from the downloaded quarter of microdata:

# one record per report
faers_demo_df <- read_faers( grep( 'DEMO23Q1\\.txt$' , unzipped_files , value = TRUE ) )

# one or more record per report
faers_drug_df <- read_faers( grep( 'DRUG23Q1\\.txt$' , unzipped_files , value = TRUE ) )

# zero or more records per report
faers_outcome_df <- read_faers( grep( 'OUTC23Q1\\.txt$' , unzipped_files , value = TRUE ) )

Construct an analysis file limited to reported deaths:

# limit the outcome file to deaths
faers_deaths_df <- subset( faers_outcome_df , outc_cod == 'DE' )

# merge demographics with each reported death
faers_df <- merge( faers_demo_df , faers_deaths_df )

# confirm that the analysis file matches the number of death outcomes
stopifnot( nrow( faers_deaths_df ) == nrow( faers_df ) )

# confirm zero reports include multiple deaths from the same reported adverse event
stopifnot( nrow( faers_df ) == length( unique( faers_df[ , 'primaryid' ] ) ) )

Save locally  

Save the object at any point:

# faers_fn <- file.path( path.expand( "~" ) , "FAERS" , "this_file.rds" )
# saveRDS( faers_df , file = faers_fn , compress = FALSE )

Load the same object:

# faers_df <- readRDS( faers_fn )

Variable Recoding

Add new columns to the data set:

faers_df <- 
    transform( 
        faers_df , 
        
        physician_reported = as.numeric( occp_cod == "MD" ) ,
        
        reporter_country_categories = 
            ifelse( reporter_country == 'US' , 'USA' ,
            ifelse( reporter_country == 'COUNTRY NOT SPECIFIED' , 'missing' ,
            ifelse( reporter_country == 'JP' , 'Japan' ,
            ifelse( reporter_country == 'UK' , 'UK' ,
            ifelse( reporter_country == 'CA' , 'Canada' ,
            ifelse( reporter_country == 'FR' , 'France' ,
                'Other' ) ) ) ) ) ) ,
        
        init_fda_year = as.numeric( substr( init_fda_dt , 1 , 4 ) )
        
    )

Analysis Examples with base R  

Unweighted Counts

Count the unweighted number of records in the table, overall and by groups:

nrow( faers_df )

table( faers_df[ , "reporter_country_categories" ] , useNA = "always" )

Descriptive Statistics

Calculate the mean (average) of a linear variable, overall and by groups:

mean( faers_df[ , "init_fda_year" ] , na.rm = TRUE )

tapply(
    faers_df[ , "init_fda_year" ] ,
    faers_df[ , "reporter_country_categories" ] ,
    mean ,
    na.rm = TRUE 
)

Calculate the distribution of a categorical variable, overall and by groups:

prop.table( table( faers_df[ , "sex" ] ) )

prop.table(
    table( faers_df[ , c( "sex" , "reporter_country_categories" ) ] ) ,
    margin = 2
)

Calculate the sum of a linear variable, overall and by groups:

sum( faers_df[ , "init_fda_year" ] , na.rm = TRUE )

tapply(
    faers_df[ , "init_fda_year" ] ,
    faers_df[ , "reporter_country_categories" ] ,
    sum ,
    na.rm = TRUE 
)

Calculate the median (50th percentile) of a linear variable, overall and by groups:

quantile( faers_df[ , "init_fda_year" ] , 0.5 , na.rm = TRUE )

tapply(
    faers_df[ , "init_fda_year" ] ,
    faers_df[ , "reporter_country_categories" ] ,
    quantile ,
    0.5 ,
    na.rm = TRUE 
)

Subsetting

Limit your data.frame to elderly persons:

sub_faers_df <- subset( faers_df , age_grp == "E" )

Calculate the mean (average) of this subset:

mean( sub_faers_df[ , "init_fda_year" ] , na.rm = TRUE )

Measures of Uncertainty

Calculate the variance, overall and by groups:

var( faers_df[ , "init_fda_year" ] , na.rm = TRUE )

tapply(
    faers_df[ , "init_fda_year" ] ,
    faers_df[ , "reporter_country_categories" ] ,
    var ,
    na.rm = TRUE 
)

Regression Models and Tests of Association

Perform a t-test:

t.test( init_fda_year ~ physician_reported , faers_df )

Perform a chi-squared test of association:

this_table <- table( faers_df[ , c( "physician_reported" , "sex" ) ] )

chisq.test( this_table )

Perform a generalized linear model:

glm_result <- 
    glm( 
        init_fda_year ~ physician_reported + sex , 
        data = faers_df
    )

summary( glm_result )

Replication Example

This example matches the death frequency counts in the OUTC23Q1.pdf file in the downloaded quarter:

stopifnot( nrow( faers_df ) == 37704 )

Analysis Examples with dplyr  

The R dplyr library offers an alternative grammar of data manipulation to base R and SQL syntax. dplyr offers many verbs, such as summarize, group_by, and mutate, the convenience of pipe-able functions, and the tidyverse style of non-standard evaluation. This vignette details the available features. As a starting point for FAERS users, this code replicates previously-presented examples:

library(dplyr)
faers_tbl <- as_tibble( faers_df )

Calculate the mean (average) of a linear variable, overall and by groups:

faers_tbl %>%
    summarize( mean = mean( init_fda_year , na.rm = TRUE ) )

faers_tbl %>%
    group_by( reporter_country_categories ) %>%
    summarize( mean = mean( init_fda_year , na.rm = TRUE ) )

Analysis Examples with data.table  

The R data.table library provides a high-performance version of base R’s data.frame with syntax and feature enhancements for ease of use, convenience and programming speed. data.table offers concise syntax: fast to type, fast to read, fast speed, memory efficiency, a careful API lifecycle management, an active community, and a rich set of features. This vignette details the available features. As a starting point for FAERS users, this code replicates previously-presented examples:

library(data.table)
faers_dt <- data.table( faers_df )

Calculate the mean (average) of a linear variable, overall and by groups:

faers_dt[ , mean( init_fda_year , na.rm = TRUE ) ]

faers_dt[ , mean( init_fda_year , na.rm = TRUE ) , by = reporter_country_categories ]

Analysis Examples with duckdb  

The R duckdb library provides an embedded analytical data management system with support for the Structured Query Language (SQL). duckdb offers a simple, feature-rich, fast, and free SQL OLAP management system. This vignette details the available features. As a starting point for FAERS users, this code replicates previously-presented examples:

library(duckdb)
con <- dbConnect( duckdb::duckdb() , dbdir = 'my-db.duckdb' )
dbWriteTable( con , 'faers' , faers_df )

Calculate the mean (average) of a linear variable, overall and by groups:

dbGetQuery( con , 'SELECT AVG( init_fda_year ) FROM faers' )

dbGetQuery(
    con ,
    'SELECT
        reporter_country_categories ,
        AVG( init_fda_year )
    FROM
        faers
    GROUP BY
        reporter_country_categories'
)