Calculate the real-world Progression-free Survival (rwPFS) endpoint

Calculates the real-world PFS (rwPFS) endpoint given information on follow-up period, progression-, and death dates.

calc_rwPFS(
  .df,
  .start_date = NULL,
  .visit_gap_start_date = NULL,
  .last_progression_abstraction_date = NULL,
  .progression_date = NULL,
  .last_activity_date = NULL,
  .death_date = NULL,
  .death_window_days = NULL,
  .max_follow_up_days = Inf,
  .label = ""
)

Arguments

.df

a data frame

.start_date

character. The name of a date column in .df that represents the baseline date

.visit_gap_start_date

character. The name of a date column in .df that represents the date of the last visit before a large (usually >90d) visit gap occurring after start_date (if any). Progression follow-up will be censored at this date (latest). Dates should be NA if there's no visit gap.

.last_progression_abstraction_date

character. The name of a date column in .df that represents the date up which real-world progression was abstracted in the database. Progression follow-up will be censored at this date (latest).

.progression_date

character. The name of a date column in .df that represents the date of first progression after start date (if any). Dates should be NA if there's no progression.

.last_activity_date

character. The name of a date column in .df that represents the date of last structural activity in the database. Progression follow-up will be censored at this date (latest).

.death_date

character. The name of a date column in .df that represents the date of death. Dates should be NA if there's no recorded death.

.death_window_days

integer. How many days after end of progression follow-up should death events be included? This is necessary because patients often drop out of the database shortly before death, and (only) death events correlated with the end of follow-up should be captured. Common values are 30-60 days, but this may depend on various factors. See vignette for how to decide on the length of this time window.

.max_follow_up_days

integer. Maximum number of days after which patients will be censored.

.label

character. Label to keep track of multiple rwPFS endpoints in one dataset.

Value

Returns a data frame with the following new data columns added:

New Columns:

rwpfs_eof_date

The end date of progression follow-up. This is the earliest non-missing date of .visit_gap_start_date, .last_progression_abstraction_date, and .last_activity_date

rwpfs_event_type

The type of rwPFS event: "Death", "Progression", "Censored", or "Missing" (if rwPFS could not be calculated)

rwpfs_date

The rwPFS event- or censoring date.

rwpfs_days

The time from .start_date to rwpfs<label>_date in days.

rwpfs_event

Whether a rwPFS event was recorded (1) or the observation was censored (0)

rwpfs_months

The time from .start_date to rwpfs<label>_date in months.

If rwPFS could not be calculated, the rwpfs<label>_event_type will be set to "Missing", and all other columns will contain missing values. This can be due to one of the following reasons: missing .last_progression_abstraction_date <= .start_date, missing .last_activity_date or <= .start_date, .last_progression_abstraction_date <= .start_date, or .death_date <= .start_date (this can happen when death dates are rounded for anonymization purposes).

Examples

if (FALSE) { # \dontrun{


library(dplyr)
library(RwPFS) 

#Starting point: the simprog simulated dataset (included in RwPFS) 

#Add rwPFS:
df <- simprog %>% 
  calc_rwPFS(
    .start_date = "baseline_date",                    #baseline date for measuring time-to-event
    .visit_gap_start_date = "visit_gap_start_date",   #the start date of any gap in visits >90 days  
    .last_progression_abstraction_date = "last_progression_abstraction_date", 
    .progression_date = "progression_date_all_events",#the date of progression (none: NA)
    .last_activity_date = "last_activity_date",       #the date of last activity in the database
    .death_date = "death_date",                       #the date of death
    .death_window_days = 30,                          #include death events <30d after progression EOF
    .max_follow_up_days = Inf,                        #censor patients after a maximum time. 
    .label = "_all_events"                            #Label for this rwPFS endpoint (for comparisons)
  )

#View the result:
df %>% 
  select(contains("rwPFS")) %>% 
  glimpse()

} # }