Page 51 - Contributed Paper Session (CPS) - Volume 7

CPS2028 Ayon M.

Covariate-Adjusted response-adaptive designs
for semi-parametric survival responses
Ayon Mukherjee
Lead Statistician, Novartis Pharmaceuticals, Hyderabad, India

Abstract
Covariate-adjusted response-adaptive (CARA) designs use the available
responses to skew the treatment allocation in a clinical trial in favour of the
treatment found at an interim stage to be best for a given patient’s covariate
profile. There has recently been extensive research on various aspects of CARA
designs with the patient responses assumed to follow a parametric model.
However, the range of application for such designs become limited in real-life
clinical trials where the responses infrequently fit a certain parametric form.
On the other hand, the parametric assumption yields robust estimates for the
covariate-adjusted treatment effects. To balance these two requirements,
designs are proposed without any distributional assumptions about the
survival responses, relying only on the assumption of proportional hazards for
the two treatment arms. To meet the multiple experimental obectives of a
clinical trial, the proposed designs are developed based on optimal allocation
approach. The optimal designs are based on biased coin procedures, with a
bias towards the better treatment arm. These are the doubly-adaptive biased
coin design (DBCD) and the efficient randomised adaptive design (ERADE). The
derived treatment allocation proportions for these designs converge to the
expected target values, which are functions of the Cox regression coefficients
that are estimated sequentially with the arrival of every new patient into the
trial. Based on simulation studies, it is found that the ERADE is preferable to
the DBCD when the main aim is to minimise the variance of the observed
allocation proportion and to maximise the power of the Wald test for a
treatment difference. However, the former procedure being discrete tends to
be slower in converging towards the expected target allocation proportion.
Other comparative merits of the proposed designs have been highlighted and
their preferred areas of application are discussed. It has been found that the
proposed designs are a suitable alternative to traditional balanced
randomisation designs in terms of their power, provided that response data
are available during the recruitment phase to enable adaptations to the
designs.

Keywords
Censored response; Optimal allocation; Power; Variability; Unbalanced
randomization.

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