Evaluating results

When evaluating the results of any meta-analysis, it is usually informative to consider the results in a number of formats. The relative difference given by the HR and (where IPD are available) illustrated by the survival curve, gives a good overall impression of the results. Absolute differences and baseline event rates give a good indication of the results specific to clinically important points in time. This is also very important when considering the results of subgroup analyses. Even where there is no evidence that there is a differential relative effect of treatment (as measured by an HR or OR) in patient subgroups, if the underlying event rates for different categories of patients differ, then the effect of treatment in absolute terms will be different. For example, if young patients have a baseline survival of 80 per cent at two years then an HR of 0.7 in favour of treatment is equivalent to a 6 per cent benefit in survival improving the survival rate from 80 to 86

Table 11.7 Baseline survival and equivalent absolute increases in survival from an IPD meta-analysis of chemotherapy in high-grade glioma. Reproduced with permission frm [36]

HR = 0.84

One year survival rate

Two year survival rate

Baseline (%)

Absolute

Baseline (%)

Absolute

increase (%)

increase (%)

Age

< 40

78

3

50

5

41-59

45

6

14

5

> 60

22

6

4

2

Sex

Male

45

6

18

5

Female

40

6

16

5

Histology

AA

58

5

31

6

GM

35

6

9

4

Other

72

4

52

5

Performance

Good

54

5

22

6

status

Poor

31

6

9

4

Extent of Resection

Complete

50

5

19

5

Incomplete

40

6

16

5

Biopsy

36

6

19

5

per cent. For old patients with a 20 per cent, two-year survival rate, the same HR of 0.7 translates to a 12 per cent benefit improving survival from 20 to 32 per cent. All of these, together with the nature of the treatment and other clinical factors should play a part in interpreting the results. It is often useful to present a table of absolute effects according to the baseline event rate for important prognostic variables. For example, Table 11.7 illustrates how the overall HR of 0.84 translates to different absolute effects ranging from 2 to 6 per cent at two years depending on the characteristics of patients with high-grade glioma.

0 0

Post a comment