Skip to main content

Table 2 Results of simulation studies

From: Augmenting the logrank test in the design of clinical trials in which non-proportional hazards of the treatment effect may be anticipated

Scenario

Dataset

n

Test

   

Cox

Joint

Perm.

Comb.

A (null)

GOG111

1000

5.2

5.3

4.9

5.3

 

PATCH1

1000

5.0

4.9

5.2

4.8

 

ICON7

1000

4.8

4.9

5.2

4.9

B (PH)

GOG111

652

9 2 . 9

87.2

8 6.7

91.0

 

PATCH1

1280

9 2 . 6

8 6.9

87.7

90.2

 

ICON7

1240

9 1 . 9

8 6.6

88.3

89.8

C (early)

GOG111

310

7 2.5

91.5

9 2 . 1

90.0

 

PATCH1

450

7 4.4

88.8

9 1 . 9

89.2

 

ICON7

522

3 6.9

9 8 . 7

92.4

89.5

D (late)

GOG111

560

92.7

9 6 . 1

8 0.5

90.3

  1. Values in table are percentages of 10,000 (scenario A) or 5,000 (scenarios B–D) simulated datasets in which each of four tests was significant at the 5 percent level. The datasets were simulated to mimic data from three randomized controlled trials with varying sample size (n). Values in bold (or italic) type indicate the most (or least) powerful among the four tests for the given scenario and dataset
  2. Abbreviations: Joint joint test [1], Perm. permutation test, Comb. combined test