Skip to main content

Table 4 Distribution of p value ratio r across 55 comparisons in 50 randomised trials

From: Combined test versus logrank/Cox test in 50 randomised trials

Ratio (r)

Evidence of non-PH

Total

r=p(CT)/p(Cox)

No (%)

Yes (%)

 

r≤0.5

2 (15)

11 (85)

13 (100)

0.5<r≤1.0

7 (64)

4 (36)

11 (100)

r>1.0

29 (94)

2 (6)

31 (100)

Total

38 (69)

17 (31)

55 (100)

  1. Based on evidence of non-PH of the treatment effect, defined by p(GT)<0.1. r<1 favours the combined test
  2. CT combined test,
  3. GT Grambsch–Therneau test,
  4. PH proportional hazards