Page 59 - Contributed Paper Session (CPS) - Volume 7

CPS2028 Ayon M.
of the DBCD design is chosen to be 2 following Sverdlov, Rosenberger and
Ryeznik (2013) .
The operating characteristics of the proposed adaptive designs as well as
the balanced randomization designs have been compared through
simulations. It has been found that, when one treatment performs better than
the other, all of the proposed CARA designs generate skewed allocations on
an average towards the better treatment arm according to covariate-specific
treatment effects. This result in fewer events in the trial without compromising
much on the statistical efficiency as compared to the balanced randomization
designs. The degree of skewness also varies according to the background
model that the design is based on. The skewness of the treatment allocation
proportions in favour of the better treatment arm thus establishes the ethical
gain of using the CARA designs as compared to the traditional balanced
randomization procedures.A slight delay in the response does not affect the
convergence of the CARA designs to their target allocation proportions. It has
been established that even without considering a distriutional assumption for
the survival responses and just assuming proportional hazard of patients with
respect to time, such ethical gain persists without heavily compromising on
the statistical power of the Wald test for the difference of the covariate-
adjusted treatment effects. It has been established by simulation results that
such ethical gain is achieved most with the DBCD based CARA design, as the
allocation proportions converge quickly to the target values. However, the
ERADE being the most efficient design provides a method that has the highest
power for testing the treatment differences. Therefore, ERADE is preferred
over the DBCD if efficiency is the sole criterion for an experimenter. However,
to keep a balance between ethics and efficiency, the DBCD procedure
outshines the ERADE and the traditional balanced randomization procedure.

5. Discussion and Conclusion
In the present paper, CARA randomization procedures for two-armed
survival trials has been considered for semi-parametric survival responses. This
means that the procedures are developed without any distriutional
assumption of the survival responses but only with a lighter assumption about
the hazards of an event at any given time point to be proportional and time-
independent for any two patients in the trial. The procedure is based on two
distinct approaches to optimality: the doubly-adaptive biased coin design and
the efficient randomised adaptive design. The relative merits of the proposed
design options, and the respective contexts favouring their preferred
applicability, have been highlighted.
The skewness of the treatment allocation proportions in favour of the
better treatment results in some reductions in the number of events in the
study and establishes the ethical gain obtained by using CARA designs. It has

46 | I S I W S C 2 0 1 9