An ongoing adaptive PII/III trial with dose selection

A pragmatic solution for a development programme

Andrew Stone 14th Dec 2007

Source: xxxxxxx

Background & rationale

AZ Case Study in an area of high unmet needDesire to develop important new medicines efficientlyWant to make sure the opportunity is taken to minimise patient exposure, and optimise resources of AZ and trialists, if the agent is insufficiently effectiveOptimise dose selection Trial design created to meet these needs– Did not know it would be called an Adaptive design BUT– Naturally only too aware of data access issues and control of

Type I error

Source: xxxxxxx

Adaptive, seamless PII/III design with dose selection

Dose1

Dose2

Control

DecisionCriteria

Met?

Phase II analysis No

Unblind data and review

YesContinue to PIII

Drop 1 dose arm

Ran

dom

ise

Source: xxxxxxx

PII data access

End of phase II (EOPII) GO criteria pre-definedAn IDMC performing the EOPII analysis– The EOPII criteria are documented in an IDMC charter

that only AZ and IDMC have access toShould the criteria be achieved, PII results will not be disclosed to either AZ or investigatorsOur philosophy is that if results are good, AZ and investigators do not need to know just how good

Source: xxxxxxx

EOPII GO criteria definition

Lack of PII access puts a premium on a thorough understanding and sponsor acceptance of GO/NO GO criteriaNecessarily a lengthy and detailed process to defineUsed predictive power methodology to guide approach– A Bayesian approach but crucially one where the prior is

based on data and not opinion

Source: xxxxxxx

Phase III data analysis

A single test of continued dose vs control group– Includes PII patients– Equal weight is given per patient– No- p-value combination or aggregating of dosesPhase III analysis approach fixed in the protocol including factors that stratify the analysis

Source: xxxxxxx

00.010.020.030.040.050.060.070.080.090.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Correlation between Zscore for PII and PIII endpoints

Type

I Er

ror

always continue

Type I error considerations for PII/III trials with dose selection

00.010.020.030.040.050.060.070.080.09

0.1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Correlation between Zscore for PII and PIII endpoints

Type

I Er

ror

always continue numerical superiority

• Final significance level requires minimal adjustment even with studies without futility analyses and a strong correlation between PII and PIII endpoints

Example where PII analysis performed at 5% of PIII events

Stallard and Todd SIM 2003 22:689-703 & Todd and Stallard DIJ 2005 39:109-118

Source: xxxxxxx

Advantages of programme

Provides the potential to deliver the new therapy to patients, in an area of high unmet need, with significant time savings Minimises patient exposure, and optimises resources AZ and trialists, if new therapy insufficiently effective

Source: xxxxxxx

Casodex 150mg (n=60)

Casodex 100mg (n=60)

Castration (n=30)

None of this new!1992 Casodex Adaptive design

Pre-planned dose selection on PSAData from patients recruited prior to dose selection included infinal efficacy assessment of Overall Survival

Dose selection

%fall in PSA

@12 wk

Casodex 150mg (n=760)

Castration (n=380)

Source: xxxxxxx

Summary

Future role of Adaptive designs at AZ– Will be considered for high potential compounds– Operationally complexIn selection situations such as this can highly be advantageous to sponsors and patients alike