CHAPTER 22 Comparing Survival Times 323
Assessing the assumptions
Like all statistical tests, the log-rank test assumes that you studied an unbiased
sample from the population about which you’re trying to draw conclusions. It also
assumes that any censoring that occurred was due to circumstances unrelated to
the treatment being tested (for example, individuals didn’t drop out of the study
because the drug made them sick).
Also, the log-rank test looks for differences in overall survival time. In other
words, it’s not good at detecting differences in shape between two survival curves
with similar overall survival time, like the two curves shown in Figure 22-4. These
two curves actually have the same median survival time, but the survival experi-
ence is different, as shown in the graph. When two survival curves cross over each
other, as shown in Figure 22-4b, the excess deaths are positive for some time
slices and negative for others. This leads them to cancel out when they’re added
up, producing a smaller z value as a test statistic z value, which translates to
larger, non-statistically significant p value.
Therefore, one very important assumption of the log-rank test is that the two
groups have proportional hazards, which means the two groups must have gener-
ally similar survival shapes, as shown in Figure 22-4a. Flip to Chapter 21 for more
about survival curves, and read about hazards in more detail in Chapter 23.
FIGURE 22-4:
Proportional (a)
and nonpropor-
tional (b) hazards
relationships
between two
survival curves.
© John Wiley & Sons, Inc.