maicplus (version 0.1.2)

maic_unanchored: Unanchored MAIC for binary and time-to-event endpoint

Description

This is a wrapper function to provide adjusted effect estimates and relevant statistics in unanchored case (i.e. there is no common comparator arm in the internal and external trial).

Usage

maic_unanchored(
  weights_object,
  ipd,
  pseudo_ipd,
  trt_ipd,
  trt_agd,
  trt_var_ipd = "ARM",
  trt_var_agd = "ARM",
  normalize_weights = FALSE,
  endpoint_type = "tte",
  endpoint_name = "Time to Event Endpoint",
  eff_measure = c("HR", "OR", "RR", "RD"),
  boot_ci_type = c("norm", "basic", "stud", "perc", "bca"),
  time_scale = "months",
  km_conf_type = "log-log",
  binary_robust_cov_type = "HC3"
)

Value

A list, contains 'descriptive' and 'inferential'

Arguments

weights_object

an object returned by estimate_weight

ipd

a data frame that meet format requirements in 'Details', individual patient data (IPD) of internal trial

pseudo_ipd

a data frame, pseudo IPD from digitized KM curve of external trial (for time-to-event endpoint) or from contingency table (for binary endpoint)

trt_ipd

a string, name of the interested investigation arm in internal trial dat_igd (real IPD)

trt_agd

a string, name of the interested investigation arm in external trial pseudo_ipd (pseudo IPD)

trt_var_ipd

a string, column name in ipd that contains the treatment assignment

trt_var_agd

a string, column name in ipd that contains the treatment assignment

normalize_weights

logical, default is FALSE. If TRUE, scaled_weights (normalized weights) in weights_object$data will be used.

endpoint_type

a string, one out of the following "binary", "tte" (time to event)

endpoint_name

a string, name of time to event endpoint, to be show in the last line of title

eff_measure

a string, "RD" (risk difference), "OR" (odds ratio), "RR" (relative risk) for a binary endpoint; "HR" for a time-to-event endpoint. By default is NULL, "OR" is used for binary case, otherwise "HR" is used.

boot_ci_type

a string, one of c("norm","basic", "stud", "perc", "bca") to select the type of bootstrap confidence interval. See boot::boot.ci for more details.

time_scale

a string, time unit of median survival time, taking a value of 'years', 'months', 'weeks' or 'days'. NOTE: it is assumed that values in TIME column of ipd and pseudo_ipd is in the unit of days

km_conf_type

a string, pass to conf.type of survfit

binary_robust_cov_type

a string to pass to argument type of sandwich::vcovHC, see possible options in the documentation of that function. Default is "HC3"

Details

For time-to-event analysis, it is required that input ipd and pseudo_ipd to have the following columns. This function is not sensitive to upper or lower case of letters in column names.

• USUBJID - character, unique subject ID

• ARM - character or factor, treatment indicator, column name does not have to be 'ARM'. User specify in trt_var_ipd and trt_var_agd

• EVENT - numeric, 1 for censored/death, 0 for otherwise

• TIME - numeric column, observation time of the EVENT; unit in days

Examples

#
# unanchored example using maic_unanchored for time-to-event data
#
data(centered_ipd_sat)
data(adtte_sat)
data(pseudo_ipd_sat)

#### derive weights
weighted_data <- estimate_weights(
  data = centered_ipd_sat,
  centered_colnames = grep("_CENTERED$", names(centered_ipd_sat)),
  start_val = 0,
  method = "BFGS"
)

weighted_data2 <- estimate_weights(
  data = centered_ipd_sat,
  centered_colnames = grep("_CENTERED$", names(centered_ipd_sat)),
  start_val = 0,
  method = "BFGS",
  n_boot_iteration = 100,
  set_seed_boot = 1234
)

# inferential result
result <- maic_unanchored(
  weights_object = weighted_data,
  ipd = adtte_sat,
  pseudo_ipd = pseudo_ipd_sat,
  trt_var_ipd = "ARM",
  trt_var_agd = "ARM",
  trt_ipd = "A",
  trt_agd = "B",
  endpoint_name = "Overall Survival",
  endpoint_type = "tte",
  eff_measure = "HR",
  time_scale = "month",
  km_conf_type = "log-log"
)
result$descriptive$summary
result$inferential$summary

result_boot <- maic_unanchored(
  weights_object = weighted_data2,
  ipd = adtte_sat,
  pseudo_ipd = pseudo_ipd_sat,
  trt_var_ipd = "ARM",
  trt_var_agd = "ARM",
  trt_ipd = "A",
  trt_agd = "B",
  endpoint_name = "Overall Survival",
  endpoint_type = "tte",
  eff_measure = "HR",
  time_scale = "month",
  km_conf_type = "log-log"
)
result$descriptive$summary
result$inferential$summary
#
# unanchored example using maic_unanchored for binary outcome
#

data(centered_ipd_sat)
data(adrs_sat)

centered_ipd_sat
centered_colnames <- grep("_CENTERED$", colnames(centered_ipd_sat), value = TRUE)
weighted_data <- estimate_weights(data = centered_ipd_sat, centered_colnames = centered_colnames)
weighted_data2 <- estimate_weights(
  data = centered_ipd_sat, centered_colnames = centered_colnames,
  n_boot_iteration = 100
)

# get dummy binary IPD
pseudo_adrs <- get_pseudo_ipd_binary(
  binary_agd = data.frame(
    ARM = rep("B", 2),
    RESPONSE = c("YES", "NO"),
    COUNT = c(280, 120)
  ),
  format = "stacked"
)

# unanchored binary MAIC, with CI based on sandwich estimator
maic_unanchored(
  weights_object = weighted_data,
  ipd = adrs_sat,
  pseudo_ipd = pseudo_adrs,
  trt_ipd = "A",
  trt_agd = "B",
  trt_var_ipd = "ARM",
  trt_var_agd = "ARM",
  endpoint_type = "binary",
  endpoint_name = "Binary Endpoint",
  eff_measure = "RR",
  # binary specific args
  binary_robust_cov_type = "HC3"
)

# unanchored binary MAIC, with bootstrapped CI
maic_unanchored(
  weights_object = weighted_data2,
  ipd = adrs_sat,
  pseudo_ipd = pseudo_adrs,
  trt_ipd = "A",
  trt_agd = "B",
  trt_var_ipd = "ARM",
  trt_var_agd = "ARM",
  endpoint_type = "binary",
  endpoint_name = "Binary Endpoint",
  eff_measure = "RR",
  # binary specific args
  binary_robust_cov_type = "HC3"
)

#---------------------------------