Median event times

We have mentioned earlier in this chapter that it is not possible to calculate the mean survival time. It is, however, usually possible to obtain median survival times Irom the Kaplan-Meier curves. The median survival time for a particular 

It is usual to estimate and plot the probability of being event-free, but there will be occasions when interpretation is clearer when the opposite of this, cumulative incidence (or cumulative probability of experiencing the event by that time), is plotted. This is simply obtained as 1 - probability of being event-free. Pocock et al. (2002) discuss issues associated with the interpretation of these plots. These authors point out that interpretation in the conventional type of plot, when the event rates are low, can be exaggerated visually by a break in the y-axis, so take care  

The Kaplan-Meier curves do not of themselves provide a formal, p-value, comparison of the treatments. This comparison of the survival curves is undertaken using either the logrank test or the Gehan-Wilcoxon test. We will look at these two test procedures in turn. 

The curves in part a) represent, over the period of follow-up, a permanent treatment effect. There is a long-term advantage in survival for one group 

One question that often arises is which test should I use as I don t know what kind of effect I am going to see My short answer to this question is that in a confirmatory setting you should know By the time you reach that stage in the drug development programme your knowledge of the disease area and the treatment, in combination with the endpoint, should enable accurate prediction of what should happen. Of course, earlier on in the programme you may not know and in this exploratory phase it is perfectly valid to undertake both tests to explore the nature of the effects. 

---

The sample size calculation should be detailed in the trial publication, indicating the estimated outcomes in each of the treatment groups (and this will define, in particular, the clinically relevant difference to be detected), the type I error, the type II error or power and, for a continuous primary outcome variable in a parallel group trial, the within-group standard deviation of that measure. For time-to-event data details on clinically relevant difference would usually be specified in terms of the either the median event times or the proportions event-free at a certain time point. 

A common measure of central tendency from the Kaplan-Meier estimate is the median survival time (note that this can be estimated only if more than half the participants experience the event). The median survival time is the earliest value of t such that the probability of survival is < 0.5. Note that when observations are censored any estimate of the mean is biased because, technically, the event would eventually occur if we followed participants indefinitely. 

Figure 15.1 shows the Kaplan-Meier (KM) estimate of the survival probability for the endpoints of PFS and OS. Median PFS time was estimated at 23.9 months and 30.0 months for placebo and treatment, respectively, and FIR = 0.82 for treatment relative to placebo with 95% Cl (0.69, 0.97) and p-value = 0.02, indicating an 18% reduction of the risk for disease progression or death. OS time was similar between two arms (FIR = 0.99 [0.79, 1.24], p = 0.93). The rates of AEl were 10% with placebo and 14.2% with active treatment, and the rates of AE2 were 0% with placebo and 6.4% with active treatment. The risk for AE2 increased with increasing exposure to treatment. We also assume that AE2 is a more serious event than AEl. It is of interest to conduct the BRA of the treatment based on these results. 

Those individuals who had whiplash pain but who did not report being depressed prior to the whiplash event were followed for future symptoms of depression, using the Epidemiological Studies Depression Scale [8j. In this follow-up, 42% of whiplash subjects who had not been previously depressed met the definition for depression over the next 6 weeks. The majority (60%) of these 42% experienced resolution within the subsequent year, with a median recovery time of 92 days, whereas 19% experienced recurrent bouts of 343... 

Of 8028 postmenopausal women with receptor-positive early breast cancer who were randomly assigned doubleblind to letrozole, tamoxifen, or a sequence of these agents for 5 years, 7963 were included in an analysis of cardiovascular events over a median follow-up time of 30 months (8). There was a similar overall incidence of cardiac adverse events (letrozole 4.8% tamoxifen4.7%), but more grade 3-5 events with letrozole (2.4% versus 1.4%), an excess that was only partly attributable to prior hypercholesterolemia. There were more thromboembolic events with tamoxifen (3.9% versus 1.7% overall and 2.3% versus 0.9% for grade 3-5 events). There were no significant differences between tamoxifen and letrozole in the incidence of hypertension or cerebrovascular events. 

The FLORIDA study recruited 540 patients with Ml and a total cholesterol level of <6,5 mmol/L and randomized them to either fluvastatin 80 mg daily or placebo at a median time of eight days after symptom onset (22). The primary composite endpoint of the study was ischemia on ambulatory electrocardiogram monitoring or a major clinical event defined as death,... 

---