correl.power: Compute Power for Failure Time Comparisons in Correlative...

In raredd/desmon: Functions for design and monitoring of clinical trials

Compute Power for Failure Time Comparisons in Correlative Studies

Description

Computes power or difference for the logrank test for a correlative study comparing failure time distributions between groups defined by two levels of a correlative marker when there is a known amount of information (number of events)

Usage

correl.power(
  n,
  nd,
  r1o2,
  p1,
  acc.per,
  add.fu,
  alpha2 = 0.05,
  int = c(0.001, 5)
)

Arguments

n	Total number of cases
nd	Total number of failures
r1o2	Ratio of the hazard in Group 1 over that in Group 2
p1	Proportion of sample in Group 1
acc.per	Duration of accrual
add.fu	Additional follow-up after accrual was completed
alpha2	Type I error rate (two-sided)
int	Initial search interval for calculating hazard rates
power	The desired power for the comparison
maxrat	The maximum hazard ratio to consider in the search when greater=TRUE. 1/maxrat is the minimum ratio to consider when less=TRUE
greater	logical; if TRUE, find the ratio greater than 1 with the specified power
less	logical; if TRUE, find the ratio less than 1 with the specified power

Details

Assumes exponential failures and a clinical trial setting with uniform accrual over acc.per units of time plus an additional add.fu time units of follow-up, so censoring is uniform on (add.fu, add.fu + acc.per). Test statistic is the exponential Wald test using asymptotic normality, which is asymptotically equivalent to the logrank test.

Only the variance under the alternative is used, consistent with Rubinstein, Gail and Santner (1981) and seqopr and srvpwr. r1o2 can be either larger or smaller than 1. The total failure probability is fixed at nd/n, and hazard rates in the two groups are computed to give the specified ratio subject to the total failure probability constraint. The uniroot function is called to solve for the hazards, and int is the initial bracketing interval used in the search. Since the magnitude of the hazard functions are dependent on the time units, it may sometimes be necessary to modify this interval.

Given a specified ratio, correl.power computes the power. Given a specified power, correl.ratio computes the corresponding hazard ratio (two ratios are computed, one for the difference in each direction).

Value

correl.power returns a vector giving the hazard rates in groups 1 and 2 (lambda1 and lambda2), the failure probabilities in the two groups (fp1 and fp2), the standard error of the log of the ratio of the estimated exponential failure rates (se), the total expected number of failures under the specified failure rates (n.fail), the normal deviate corresponding to the type II error (zbeta), and the power (power).

correl.ratio returns a vector giving the total number of failures (n.fail), the proportion of the sample expected to be in group 1 (pr.g1), and the hazard ratios (group 1/group 2) for which the test will have the specified power in the directions with worse and better outcome on group 1 (r1o2.1worse and r1o2.1better).

References

Rubinstein, Gail, Santner (1981). J Chron Dis 8: 67-74. Rubinstein, Gail, Santner (1981). J Chron Dis 34: 469-479.

See Also

powlgrnk

Examples

correl.power(355,172,1.54,.4,7,4)
correl.power(355,172,3.36,.92,7,4)
correl.power(355,172,1.67,.7,7,4)

# check symmetry
correl.power(355,172,2/3,.5,7,4)
correl.power(355,172,1.5,.5,7,4)
correl.power(355,172,2/3,.25,7,4)
correl.power(355,172,1.5,.75,7,4)

correl.ratio(1000,150,.2,2,3.635,int=c(.0001,20))

raredd/desmon documentation built on May 7, 2024, 3:46 p.m.