Just How Flexible Can a Prospective Clinical Trial Be?

Just How Flexible Can a Prospective Clinical Trial Be?

Donald A. Berrydberry@mdanderson.org

Donald A. Berrydberry@mdanderson.org

BERRY

STATISTICAL INNOVATION

CONSULTANTS

22

Definition of “flexible”Definition of “flexible”

Pliant Characterized by a ready

capability to adapt to new, different, or changing requirements

Pliant Characterized by a ready

capability to adapt to new, different, or changing requirements

33

With thanks to Mike Krams & Vlad DragalinWith thanks to Mike Krams & Vlad Dragalin

44

Original question is not helpful

Original question is not helpful

Two kinds of flexibility: Ad hoc (improvised) Planned (prospective)

Two kinds of flexibility: Ad hoc (improvised) Planned (prospective)

55

Original question is not helpful

Original question is not helpful

Two kinds of flexibility: Ad hoc: okay for personal or

internal company decisions Planned: essential for

external buy-in

Two kinds of flexibility: Ad hoc: okay for personal or

internal company decisions Planned: essential for

external buy-in

66

New question 1: How to deal with ad hoc flexibility

in a clinical trial?

New question 1: How to deal with ad hoc flexibility

in a clinical trial?

You can’t!You can’t!

77

New question 2: How to deal with planned flexibility

in a clinical trial?

New question 2: How to deal with planned flexibility

in a clinical trial?

Design

The team on board?

Logistics

Regulatory

Design

The team on board?

Logistics

Regulatory

88

DesignDesign

Modeling--arbitrarily complicated Efficacy and safety Computation OCs & comparisons

Type I error Power Sample size Alternative designs

Modeling--arbitrarily complicated Efficacy and safety Computation OCs & comparisons

Type I error Power Sample size Alternative designs

99

The team on board?The team on board? You have to teach them

And you have to keep teaching them!

“I don’t want the DSMB making decisions such as whether to go to phase III.”

DSMB as automaton

You have to teach them

And you have to keep teaching them!

“I don’t want the DSMB making decisions such as whether to go to phase III.”

DSMB as automaton

1010

LogisticsLogistics Data flow: Updating database Clean data? (no, but model)

QA Auxiliary variables Central review

Missing data--same as above What CRO? Information leakage--who

knows what and when?

Data flow: Updating database Clean data? (no, but model)

QA Auxiliary variables Central review

Missing data--same as above What CRO? Information leakage--who

knows what and when?

1111

Logistics (cont’d)Logistics (cont’d)

“Adaptive design is a whole new ball game. If you don’t know what you are doing, you can do some very bad things and damage the integrity of the trial, and of the whole process.”

“Adaptive design is a whole new ball game. If you don’t know what you are doing, you can do some very bad things and damage the integrity of the trial, and of the whole process.”

1212

Who knows what & when?Who knows what & when? Phase I or II vs Phase III Dose-finding (“The Pharmacist Knows”) Dropping arms or adaptive

randomization (blinded vs open) Seamless phase II/III

Simple line in the sand (Publish but continue to accrue?)

“End of phase II meeting” Expanding accrual One study or two?

Phase I or II vs Phase III Dose-finding (“The Pharmacist Knows”) Dropping arms or adaptive

randomization (blinded vs open) Seamless phase II/III

Simple line in the sand (Publish but continue to accrue?)

“End of phase II meeting” Expanding accrual One study or two?

1313

RegulatoryRegulatory

CRCs and IRBs

“Please explain in lay terms” (Examples, examples, examples)

Informed consent?

FDA, EMEA

Meet early & as often as possible

CRCs and IRBs

“Please explain in lay terms” (Examples, examples, examples)

Informed consent?

FDA, EMEA

Meet early & as often as possible

1414

The Bottom LineThe Bottom Line

Clinical trials will continue to become more complicated (such as I-SPY2)

This is good and bad Trials must begin to address

interactions between treatments and patient covariates

Clinical trials will continue to become more complicated (such as I-SPY2)

This is good and bad Trials must begin to address

interactions between treatments and patient covariates

1515

The Bottom LineThe Bottom Line Traditional trials are poorly suited

for identifying treatments that benefit patients who have hetero-geneous diseases such as cancer

The adaptive path is treacherous, with mines every step of the way

Avoiding mines is challenging, but enormously rewarding ... and fun!

Traditional trials are poorly suited for identifying treatments that benefit patients who have hetero-geneous diseases such as cancer

The adaptive path is treacherous, with mines every step of the way

Avoiding mines is challenging, but enormously rewarding ... and fun!