Advanced Prostate Cancer: A Meta-Analysis of Two Trials

This dataset comprises two trials that compared oral liarozole, an experimental retinoic acid metabolism-blocking agent (Z = 1), with an antiandrogenic drug considered as control (Z = 0): cyproterone acetate in the first trial and flutamide in the second (Buyse et al., 2003). In both trials, patients were in relapse after first-line endocrine therapy. The trials accrued 312 and 284 patients, respectively. Each trial was multinational and multicentric, and the unit of analysis for the surrogacy analysis was chosen to be the country in which the patients were treated. There were 19 countries containing between 4 and 69 patients. The primary endpoint of the trials was overall survival from the start of treatment. In both trials, patients were assessed at baseline (before the start of treatment), at 2 weeks, monthly for 6 months, at 3-month intervals until the second year, and at 6-month intervals until treatment discontinuation or death. The assessments included measurement of the prostate-specific antigen (PSA) level. PSA is a glycoprotein that is found almost exclusively in normal and neoplastic prostate cells. Changes in PSA often antedate changes in bone scan, and they have been used as an indicator of response in patients with androgen-independent prostate cancer.

For the surrogacy analyses, overall survival will be considered the true

TABLE 2.9

Advanced Prostate Cancer Trials. Overview of the variables in the dataset.

endpoint (T), with the full longitudinal PSA profile of each patient considered the surrogate (Buyse et al., 2003; Renard et al., 2003). In particular, the logarithm of PSA measurements will be used.

Figure 2.7 presents survival curves per treatment arm for both trials. Note that there were 18 patients without PSA measurements, which will be excluded from the analyses using PSA measurements.

Figure 2.8 presents the scatter plots of the observed log-PSA values in function of the measurement time for patients per trial. Additionally, the plots include the smoothed mean profiles for the control and treatment groups.

In the remainder of this book, this dataset will be referred to as the PSA dataset. The dataset is included in the R library Surrogate (where it can be accessed using the command data(PSA) and it can be downloaded from http://ibiostat.be/online-resources (file PSA.txt). Table 2.9 provides an overview of the variables that are included in the PSA dataset. The data were provided by the Janssen Research Foundation (see Buyse et al., 2003).

Acknowledgments for Use of Data

The example datasets that accompany this book can be freely downloaded from http://ibiostat.be/online-resources. Researchers who download and analyze the data accept the following conditions:

• 1. the research shall be scientifically sound and peer reviewed,
• 2. the data shall be used for the purpose of evaluating surrogacy,
• 3. the source of the data shall be acknowledged in all presentations and publications, including a reference to publication(s) by the investigator or group of investigators who generated the data,

• 4. the results of the analyses shall be made available to the research community, and
• 5. the confidentiality of individual patient data shall be protected.