Introduction to Bayesian methods III: use and interpretation of Bayesian tools in design and analysis

Abstract

The Bayesian approach and several of its advantages in drug and medical device development are described. One advantage from the perspective of analysis is that it provides a methodology for synthesizing information. However, taking a Bayesian approach to designing clinical trials is potentially more valuable than using this approach in analyzing trial results. Bayesian methodology provides a mechanism for updating what is known as results accumulate during a trial. Such updating can be incorporated completely explicitly and prospectively. An important way in which the Bayesian approach can be used is in calculating the predictive probability distribution of future results on the basis of current results. I show how to exploit predictive distributions in adapting to results that accumulate during the course of a trial. Possible adaptations including decreasing or increasing sample size, dropping treatment arms, and modifying the randomization proportions to the various arms depending on the interim results. Consequences of taking a Bayesian approach to clinical trial design are efficiency, better treatment of patients in the trial, and greater precision regarding the primary endpoints. An example of the last of these is Bayesian modeling of the relationship between early and longer term endpoints. Such modeling also enables earlier decision making. Case studies 2 and 3 deal with trials that were shorter and smaller, respectively, because of such modeling.

Get full access to this article

View all access options for this article.

References

1. Buzdar AU , Ibrahim NK , Francis D et al. Significantly higher pathological complete remission rate following neoadjuvant therapy with trastuzumab, paclitaxel and epirubicin-containing chemotherapy: results of a randomized trial in HER-2-positive operable breast cancer. J Clin Oncol 2005; 23: 3676–3685.

2. Giles FJ , Kantarjian HM , Cortes JE et al. Adaptive randomized study of idarubicin and cytarabine versus troxacitabine and cytarabine versus troxacitabine and idarubicin in untreated patients 50 years or older with adverse karyotype acute myeloid leukemia. J Clin Oncol 2003; 21: 1722–1727.

3. Berry DA , Müller P , Grieve AP et al. Adaptive Bayesian designs for dose-ranging drug trials. In Gatsonis C , Carlin B , Carriquiry A , eds. Case studies in Bayesian statistics V., New York: Springer-Verlag,2001, 88–181.

4. Inoue LYT , Thall P , Berry DA. Seamlessly expanding a randomized phase II trial to phase III. Biometrics 2002; 58: 264–272.

5. Retrieved 7 June 2005 from: http://www.fda.gov/ohrms/dockets/ac/02/slides/3829s2_03_Bristol-Meyers-metaanalysis.ppt

6. Berry SM , Berry DA , Natarajan K , Lin C-S , Hennekens CH , Belder R. Bayesian survival analysis with nonproportional hazards: Metaanalysis of pravastatin-aspirin. J Am Stat Assoc 2004, 99: 36–44.

7. Berry DA . Statistical innovations in cancer research. In Kufe DW , Pollock RE , Weichselbaum RR , Bast RC , Gansler TS , Holland JF , Frei E III , eds. Cancer medicine, sixth edition. London: BC Decker, 2003, 465–478.

8. Berry DA. Bayesian statistics and the efficiency and ethics of clinical trials. Statistical Science 2004; 19: 175–187.

9. Couzin J. The new math of clinical trials. Science 2004; 303: 784–786.

10.

10. Berry DA. Adaptive clinical trials and Bayesian statistics (with discussion). Pharmaceutical Report 2002; 9: 1–11.

11.

11. Berry SM , Berry DA. Accounting for multiplicities in assessing drug safety: A three-level hierarchical mixed model. Biometrics 2004, 60: 418–426.