Table of AE summary by body system class

aetable is used to plot a table of AE summary by body system class and arm using the following summary statistics:

Total number of participants at risk per arm
Frequency (number of participants with at least one event for each body system class)
Proportions (number of participants with at least one event for each body system class relative to number of participants at risk)
Number of adverse events per participant: presented as counts, mean (SD)
Total number of events and incidence rates (number of events relative to total time in follow-up)
Treatment effect estimate and 95% confidence intervals

Usage

aetable(
  data,
  control,
  intervention_levels,
  body_system_class = "body_system_class",
  id = "id",
  arm = "arm",
  date_rand = "date_rand",
  last_visit = "last_visit",
  control_name = NULL,
  intervention_names = NULL,
  treatment_effect_estimate = TRUE,
  model = "Poisson (rate)",
  variables = c(),
  mean = TRUE,
  drop_bodsys = TRUE,
  IR_per_person = 100,
  proportions_dp = 1,
  IR_dp = 1,
  mean_dp = 1,
  SD_dp = 1,
  estimate_sf = 3,
  CI_sf = 3,
  save_image_path = NULL,
  save_docx_path = NULL
)

Arguments

data: data frame with adverse_event, body_system_class, id, arm, date_rand and last_visit columns (optional: variables to be included in model)
control: factor level of control arm
intervention_levels: vector of factor levels for interventions
body_system_class: name of body_system_class column
id: name of id column
arm: name of arm column
date_rand: name of date_rand column
last_visit: name of last_visit column
control_name: name of control arm
intervention_names: vector of names for interventions
treatment_effect_estimate: a logical value whether to include treatment effect estimate and 95% CI column in summary table (only for 2 arms)
model: model used for computation of treatment effect estimate and 95% CI
variables: vector of variable names to be included in the model for computation of treatment effect estimate (excluding arm)
mean: a logical value whether to include mean and SD column in summary table
drop_bodsys: a logical value whether to drop body system class with no observations from summary table
IR_per_person: incidence rate per number of person
proportions_dp: number of decimal places for proportions
IR_dp: number of decimal places for incidence rate
mean_dp: number of decimal places for mean number of AEs per participant
SD_dp: number of decimal places for standard deviation of number of AEs per participant
estimate_sf: number of significant figures for treatment effect estimate
CI_sf: number of significant figures for 95% CI
save_image_path: file path to save table as image
save_docx_path: file path to save table as docx

Value

flextable of AE summary by body system class and arm

Treatment Effect Estimate

The treatment effect and its 95% confidence interval are estimated via the following models below:

Poisson (rate): fitting a generalised linear model with Poisson family and log link with length of follow up time as offset
Poisson (count): fitting a generalised linear model with Poisson family and log link with count as response
Negative Binomial (rate): fitting a Negative Binomial model with length of follow up time as offset
Negative Binomial (count): fitting a Negative Binomial model with count as response
Binomial (logit): fitting a generalised linear model with Binomial family and logit link
Binomial (log): fitting a generalised linear model with Binomial family and log link
Binomial (identity): fitting a generalised linear model with Binomial family and identity link

The default model is the Poisson (rate) model.

For Poisson and Negative Binomial models, the treatment effect estimate is the Incidence Rate Ratio (IRR). For Binomial (logit) model, the treatment effect estimate is the Odds Ratio (OR). For Binomial (log) model, the treatment effect estimate is the Risk Ratio (RR). For Binomial (identity) model, the treatment effect estimate is the Risk Difference (RD).

The reference group for the treatment arm in the regression model is the control arm.

Additional covariates besides arm can be added into the model via the argument variables. Note that interaction terms cannot be added to the model.

Examples

library(flextable)
set_flextable_defaults(font.family="sans")
df2$aebodsys <- as.factor(df2$aebodsys)
aetable(df2, body_system_class="aebodsys", control="Placebo", intervention_levels=c("Intervention"), treatment_effect_estimate=TRUE, variables = c("variable1", "variable2"))