Zero early weight for weighted logrank tests
Arguments
- x
A
counting_process()-class data frame with a counting process dataset.- early_period
The initial delay period where weights increase; after this, weights are constant at the final weight in the delay period.
- fail_rate
A data frame record the failure rate.
Value
A data frame. The column weight contains the weights for the
early zero weighted logrank test for the data in x.
References
Xu, Z., Zhen, B., Park, Y., & Zhu, B. (2017). "Designing therapeutic cancer vaccine trials with delayed treatment effect." Statistics in Medicine, 36(4), 592--605.
Examples
library(dplyr)
#>
#> Attaching package: ‘dplyr’
#> The following objects are masked from ‘package:stats’:
#>
#> filter, lag
#> The following objects are masked from ‘package:base’:
#>
#> intersect, setdiff, setequal, union
library(gsDesign2)
# Example 1: Unstratified
sim_pw_surv(n = 200) |>
cut_data_by_event(125) |>
counting_process(arm = "experimental") |>
early_zero_weight(early_period = 2) |>
filter(row_number() %in% seq(5, 200, 40))
#> stratum events n_event_tol tte n_risk_tol n_risk_trt s
#> 1 All 1 0 0.2412922 196 99 0.9800000
#> 2 All 1 0 2.9852132 154 77 0.7782867
#> 3 All 1 0 9.2281147 113 63 0.5751022
#> 4 All 1 0 26.4142469 9 7 0.2751487
#> o_minus_e var_o_minus_e weight
#> 1 -0.5051020 0.2499740 0
#> 2 -0.5000000 0.2500000 1
#> 3 -0.5575221 0.2466912 1
#> 4 -0.7777778 0.1728395 1
# Example 2: Stratified
n <- 500
# Two strata
stratum <- c("Biomarker-positive", "Biomarker-negative")
prevalence_ratio <- c(0.6, 0.4)
# Enrollment rate
enroll_rate <- define_enroll_rate(
stratum = rep(stratum, each = 2),
duration = c(2, 10, 2, 10),
rate = c(c(1, 4) * prevalence_ratio[1], c(1, 4) * prevalence_ratio[2])
)
enroll_rate$rate <- enroll_rate$rate * n / sum(enroll_rate$duration * enroll_rate$rate)
# Failure rate
med_pos <- 10 # Median of the biomarker positive population
med_neg <- 8 # Median of the biomarker negative population
hr_pos <- c(1, 0.7) # Hazard ratio of the biomarker positive population
hr_neg <- c(1, 0.8) # Hazard ratio of the biomarker negative population
fail_rate <- define_fail_rate(
stratum = rep(stratum, each = 2),
duration = c(3, 1000, 4, 1000),
fail_rate = c(log(2) / c(med_pos, med_pos, med_neg, med_neg)),
hr = c(hr_pos, hr_neg),
dropout_rate = 0.01
)
# Simulate data
temp <- to_sim_pw_surv(fail_rate) # Convert the failure rate
set.seed(2023)
sim_pw_surv(
n = n, # Sample size
# Stratified design with prevalence ratio of 6:4
stratum = tibble(stratum = stratum, p = prevalence_ratio),
# Randomization ratio
block = c("control", "control", "experimental", "experimental"),
enroll_rate = enroll_rate, # Enrollment rate
fail_rate = temp$fail_rate, # Failure rate
dropout_rate = temp$dropout_rate # Dropout rate
) |>
cut_data_by_event(125) |>
counting_process(arm = "experimental") |>
early_zero_weight(early_period = 2, fail_rate = fail_rate) |>
filter(row_number() %in% seq(5, 200, 40))
#> stratum events n_event_tol tte n_risk_tol n_risk_trt
#> 1 Biomarker-negative 1 0 0.5257917 142 74
#> 2 Biomarker-negative 1 0 4.1066618 79 43
#> 3 Biomarker-positive 1 1 1.9657427 153 74
#> 4 Biomarker-positive 1 0 10.7059605 21 10
#> s o_minus_e var_o_minus_e hr duration weight
#> 1 0.9731944 -0.5211268 0.2495537 0.8 4 0.0
#> 2 0.6800393 -0.5443038 0.2480372 0.8 4 0.8
#> 3 0.8822199 0.5163399 0.2497330 0.7 3 0.0
#> 4 0.5265625 -0.4761905 0.2494331 0.7 3 0.7