size_two_arm: Sample size

In npsurvSS: Sample Size and Power Calculation for Common Non-Parametric Tests in Survival Analysis

Sample size

Description

Calculate required sample size and expected number of events for a two-arm survival study.

Usage

size_two_arm(
  arm0,
  arm1,
  test = list(test = "weighted logrank"),
  power = 0.8,
  alpha = 0.025,
  sides = 1
)

Arguments

arm0	object of class 'arm'.
arm1	object of class 'arm'.
test	list or list of lists. Each list must contain at minimum the key 'test' describing the type of statistical test. Default test is the "weighted logrank". Kaplan-Meier based tests ("survival difference", "survival ratio", "rmst difference", "rmst ratio", "percentile difference", and "percentile ratio") require the user to define an additional key, either the desired 'milestone' or 'percentile'. The weighted log-rank test does not require additional keys. However, user may choose which weight function ("1"=unweighted, "n"=Gehan-Breslow, "sqrtN"=Tarone-Ware, "FH_[a]_[b]"= Fleming-Harrington with p=a and q=b) and which approximation for the large-sample mean ("asymptotic", "generalized schoenfeld", "event driven", "freedman", "rubinstein") and variance ("1", "block[ randomization]", "simple[ randomization]") they wish to use. Default choice is 'weight'="1", 'mean.approx'="asymptotic", and 'var.approx'="1". For more details regarding the different mean and variance approximations for the weight log-rank test, please see Yung and Liu (2020). If there are multiple lists, then users may provide a 'label' for each list to be displayed in the output.
power	1 - type 2 error rate
alpha	type 1 error rate
sides	1=1-sided test, 2=2-sided test

Value

n0	sample size for arm0
n1	sample size for arm1
n	total sample size
d0	expected number of events for arm0
d1	expected number of events for arm1
d	total expected number of events; can be used to convert a time-driven trial to an event-driven trial.

References

Yung, G and Liu, Y. (2020). Sample size and power for the weighted log-rank test and Kaplan-Meier based tests with allowance for non-proportional hazards. Biometrics 76(3):939-950.

See Also

create_arm for creating an object of class 'arm'.

Examples

arm0 <- create_arm(size=120, accr_time=6, surv_scale=0.05, loss_scale=0.005, follow_time=12)
arm1 <- create_arm(size=120, accr_time=6, surv_scale=0.03, loss_scale=0.005, follow_time=12)
size_two_arm(arm0, arm1)
size_two_arm(arm0, arm1, list(test="weighted logrank",
  weight="n",
  mean.approx="generalized schoenfeld",
  var.approx="block"))
size_two_arm(arm0, arm1, list(test="survival difference", milestone=12))
size_two_arm(arm0, arm1, list(test="rmst ratio", milestone=12))
size_two_arm(arm0, arm1, list(test="percentile difference", percentile=0.25))
size_two_arm(arm0, arm1, list(
  list(test="weighted logrank", label="Logrank"),
  list(test="survival difference", milestone=12, label="12-month survival difference")))

npsurvSS documentation built on May 29, 2024, 11:23 a.m.