Chapter 4
Borrowing Strength with
Hierarchical Models over
Non-Exchangeable Subpopulations
4.1 Outline
We address statistical inference in a phase II trial of sarcoma. Sarcoma is a rare can-
cer affecting connective and soft tissues (e.g., cartilage and fat). Sarcoma is a very
heterogeneous disease with many different subtypes with widely different prognosis.
In the proposed design we pay particular attention to the heterogeneous nature of the
disease and the fact that different disease subtypes are related but can not be con-
sidered a priori exchangeable. We classify different subtypes by the overall prognosis
as poor, intermediate or good. The objective of the study is to assess the efficacy
of a new drug in patients with different subtypes of sarcoma. Let pi, i = 1,... ,n,
denote the probability of response (defined below) for a patient with disease subtype
i. One possible approach is to analyze the different subtypes as separate studies.
But due to the rare nature of some of the subtypes the enrollment in n — 12 sepa-
rate studies would be way too slow. This leads us to consider borrowing of strength
81