Bias Assessment in Progression-Free Survival Analysis Using Interval-Censored Methods - Interval-Censored-Time-to-Event Data: Methods and Applications - page 292

TABLE 10.7: Type I Error Rate (%) at = 5% (two-sided): 60% Event

Rate, 12 Weeks Median in C

Scenario Interval Right-point Mid-point Finkelstein's Sun's

6 5.4 5.4 5.6 5.5

I 8 5.4 5.4 5.2 5.2

12 5.6 5.6 5.6 5.3

6 5.8 5.5 5.6 5.3

II 8 5.4 5.3 5.6 5.2

12 6.3 5.2 5.8 5.4

6 5.4 5.1 5.4 5.3

III 8 5.4 5.5 5.6 5.5

12 5.6 4.6 5.7 5.7

6 16.6 16.6 5.8 5.4

IV 8 27.8 27.8 5.8 5.1

12 62.1 62.1 6.0 5.4

6 14.1 14.1 4.6 4.5

V 8 25.4 25.4 4.7 4.8

12 60.3 60.3 5.5 4.9

ExactlogranktestTypeIerror5.2%.

Empirical power rates at = 5% (two-sided) with dierent lengths of me-

dian PFS and event rates are analogously summarized in Tables 10.10 through

10.13. We observe consistent results as point estimates across all scenarios, and

interval-censored tests remain robust under different settings and scenarios.

In scenario I (per-protocol compliance), we find comparable power rates with

both conventional and interval-censored methods. When compared with the

logrank tests of exact failure time, it is clearly seen that, as assessment fre-
quency and event proportion decrease, all tests become less powerful. These

results are expected as less information contained in observed data leads to

less powerful tests. Similar results also hold true under Scenarios II and III.