Table of Contents:

Analysis of a Case Study

In this section, we illustrate the application of the methods, mentioned in the previous section, by the individual-patient-data meta-analysis of patients with advanced (metastatic) prostate cancer (see Section 2.2.7). The goal of the analysis is to evaluate the repeated measurements of (the logarithm of) PSA as a surrogate for overall survival (OS). The data were analyzed before by Renard et al. (2003) and Buyse et al. (2003). For analysis purposes, patients were grouped by trial and by country (see also Section 7.3). Treatment with flutamide or cyproterone acetate is considered as the control treatment, while liarozole is regarded as the experimental treatment.

Using SAS

The dataset includes PSA measurements for 596 patients. There are 19 trial- by-country groups containing between four and 69 patients per group. Two groups (one with four and one with eight patients) have to be eliminated from the analysis, because at least one of the treatment arms within the group does not contain any deaths. Consequently, the analysis includes 567 patients spread across 17 groups (see Table 8.1).

Figure 2.8 presents the scatter plots of the observed log-PSA values as a function of the measurement time for patients per trial. Additionally, the plots include the smoothed profiles for the control and treatment groups. It can be seen that most of the patients had their PSA measurements taken within the first few months after randomization. Also, it is worth noting that the smoothed profiles are not fully informative about the change in the log- PSA levels over time due to drop-out. In particular, patients who experienced a PSA progression, i.e., an increase in the PSA level, left the study. Hence, the smoothed profiles, especially at the later points in time, do not include those patients. A more informative way of looking at log-PSA levels over time is to consider cohorts of patients defined by the time they leave the study (for


Prostate Cancer. Smoothed mean log-PSA profiles for cohorts of patients with similar follow-up times. Solid line for the control treatment, dashed line for the experimental one.

any reason). Smoothed profiles for such cohorts (with observation time in the interval of (0, 0.25], (0.25, 0.5], (0.5, 0.75], (0.75, 1], (1, 1.25], and (1.25, 1.5] years) are shown in Figure 8.1. Note that the plot does not include profiles of patients with follow-up longer than 1.5 years, as there were too few such patients. The smoothed profiles in Figure 8.1 indicate the tendency of PSA to go down initially (PSA response) and then increase again (PSA progression).

< Prev   CONTENTS   Source   Next >