Improving Accrual to Clinical Trials Saving patients and $aving Time Daniel Weisdorf October 2011 Thanks to Mary Horowitz.

Improving Accrual to Clinical Improving Accrual to Clinical Trials Trials

Saving patients and $aving Saving patients and $aving TimeTime

Daniel Weisdorf

October 2011

Thanks to Mary Horowitz

Clinical ResearchClinical Research

Basic Biomedical Research

Human Studies of Safety/Efficacy –

Clinical Trials

Clinical Science

and Knowledge

Clinical Practice and Health Decision

Making

Improved Health

From: Sung et al. Central challenges facing the national clinical research enterprise. JAMA 2003;289:1278-87.

Effectiveness vs. Efficacy

Quality/Access

Improvement

ImpairedHealth

Major frequent problems in Clinical Trials

1. Bad idea

2. Poor study design--masks true test of a good idea

competing hazards, co-variates, wrong patient group

3. Too few patients--statistically invalid.

May not observe great response in subset.

4. Too short a trial.

5. Wrong analysis. ,ß...….. errors

CLINICAL TRIALS ARE DIFFICULTCLINICAL TRIALS ARE DIFFICULT

2003 Institute of Med Report High research costs/lack of funding Regulatory burden Fragmented infrastructure/ incompatible databases Lack of willing participants (informed consent issues+

) Lack of qualified investigators Career disincentives

True for all disciplines – but especially in HCT

Distribution of Transplant Volumes Distribution of Transplant Volumes among 181 US Centers Reporting Data to among 181 US Centers Reporting Data to

CIBMTR-2008CIBMTR-2008

34%

15%22%

25%

4%

<3030-4950-99100-299>300

Mmh06_6.ppt

29% >100/y

HCT is really uncommon

Distribution of Distribution of AllogeneicAllogeneic Transplant Volumes Transplant Volumes

among 181 US Centers Reporting Data to among 181 US Centers Reporting Data to CIBMTR -2008CIBMTR -2008

61%16%

15%

7% 1%

<3030-4950-99100-299>300

Mmh06_6.ppt

Individual centers treat relatively few heterogeneous patients. Many patient & treatment factor affect outcomes.

23% >100/y

Allogeneic HCT is even more uncommon

ALLOGENEIC TRANSPLANTS ELIGIBLE FOR A TRIAL OF AN INTERVENTION FOR AML-CR1

For Phase III trial needing 200 patients over 2 years, 25% of all eligible patients must be enrolled – Most definitive trials will require multicenter

participation

7,000

2,0005,000

AMLOther disease

8001200

HLA-ident sibOther donor

400

in ~200 centers

400

CR1Not CR1

HCT Trials are difficultHCT Trials are difficult

Complex therapy for life-threatening diseases Long, involved informed scary consent documents Patients are very sick

Everyone [patient/referring MD/transplant MD] wants to believe they know the best treatment

Competing risks complicate the primary endpoint How do you interpret patients who die before reach a

primary endpoint? Death before engraftment or GVHD

Need more patients to make up for drop-out Two subjects – donor and recipient

Increases complexity and costs Decreases numbers eligible for studies

BMT endpoints occur at varying times

Engraftment at 20-45 days;

GVHD by 100+ days; Chronic GVHD 2-24 months

Relapse 1-2 years;

Survival long term.

Quit for early endpoint likely makes sample too small to see

late endpoint.

Followup is costly but well designed early trial can be

informative later on

DBV06_21.ppt

Small NumbersSmall Numbers Tests our willingness to collaborate.

Large collaborative observational databases – a unique feature of our field CIBMTR EBMT

Usable for study planning—control and eligible pts

Collaboration for clinical trials

– 2001 – Blood and Marrow Transplant Clinical Trials Network BMT CTN – infrastructure for large Phase II and III trials

– 2005 – Resource for Clinical Investigations in BMT RCI BMT – smaller Phase I/II trials

What makes a good Protocol?What makes a good Protocol? Important question

Can the result change practice? How many people might it affect?

What is a realistic effect size? How to estimate baseline outcomes

and treatment effects Affects target enrollment

How many patients are there that might be eligible to enroll? Who is treating them now and how?



What is a realistic effect size? What are realistic baseline outcomes

and treatment effects Determines target enrollment




What is a realistic effect size? How to estimate baseline outcomes and

treatment effects Affects target enrollment

How many patients are there eligible to enroll? Who is treating them now; and how? Are they interested in a new HCT trial?

Baseline Outcomes and Effect SizesBaseline Outcomes and Effect Sizes

Publication bias – results may be misleading Single center study – examine your own

data Multicenter study – CIBMTR database

TCP99_30.ppt

95% CONFIDENCE INTERVALS FOR STUDIES PRODUCING 50% SURVIVAL

SAMPLE SIZE, N

0 20 30 40 50 60 70 80 90 10010

PR

OB

AB

ILIT

Y,

%

0

50

100

300200

Publish

+

Don’t

Publish +

most papers

TCP99_30.ppt

SAMPLE SIZE, N

0 20 30 40 50 60 70 80 90 10010

PR

OB

AB

ILIT

Y,

%

0

50

100

300200

Publish

+

Don’t

Publish +Most trials

95% CONFIDENCE INTERVALS FOR STUDIES PRODUCING 50% SURVIVAL

Not all have equipoise

Baseline Outcomes and Effect SizesBaseline Outcomes and Effect Sizes

Publication bias – results may be misleading Single center study – examine your own data Multicenter study – CIBMTR/EBMT databases

Consider randomized Phase II study

Allows consideration of broader eligiblity and selection to judge worthiness for followup study

Can use pick the winner design-not the best, but the best to study

Randomized Phase II TrialsRandomized Phase II Trials Single arm Phase II designs assume historical rate is

“fixed” and efficacy based on a one sample comparison to that a fixed rate

A randomized control arm may be helpful if historical rate is uncertain Especially useful when inclusion criteria are variable:

VOD syndrome, high risk GVHD

BMT CTN 0302 – randomized 4 arm Phase II of treatment for high-risk acute GVHD Response rate in eligible population was not well

understood—and 4 agents were promising

Cautions when you write the Cautions when you write the protocolprotocol

Treatment regimen – How closely does it mimic practice? Is it unnecessarily complex?

Supportive care Fussy (single institution habit)

requirements can prevent centers from participating

Too many restrictions can prevent patients from participating

Treatment regimen – How closely does it mimic practice? Is it unnecessarily complex?

Supportive care Fussy (single institution habit)

requirements can prevent centers from participating

Too many restrictions can prevent patients from participating

Collect only what you need forSafety; analysis; planning next trial

CautionsCautions when you write the when you write the protocolprotocol

BMT CTN Protocol 0302 BMT CTN Protocol 0302 Randomized Phase II Trial of Combined Randomized Phase II Trial of Combined

Therapy with Prednisone with Either MMF, Therapy with Prednisone with Either MMF, Etanercept, Pentostatin or Ontak for Newly Etanercept, Pentostatin or Ontak for Newly

Diagnosed AGVHDDiagnosed AGVHD

020406080

100120140160180200

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11

Actual

RevisedProjection

OriginalProjection

Looser eligibilityEasier start 72 not 48hWider Pred dose range

Writing a good protocolWriting a good protocol


Approximate Annual Number of US Approximate Annual Number of US Transplants Fulfilling Eligibility Criteria Transplants Fulfilling Eligibility Criteria

for BMT CTN Trialsfor BMT CTN Trials

0

500

1000

1500

2000

2500Not Enrolled Enrolled

Rapid accrual

% of Eligible US Transplants % of Eligible US Transplants Enrolled on BMT CTN TrialsEnrolled on BMT CTN Trials

0%

20%

40%

60%

80%

100%

#010

2

#010

1

#020

1

#040

2

#030

3

#040

1

#050

1

#040

3

#030

1

#060

4

#060

3

#070

1

#060

1

Not Enrolled Enrolled

Unique trials

Understanding the Potential Patient Understanding the Potential Patient PopulationPopulation

Is the study feasible? Does accrual require capture of 10% [or

50%] of all transplants Number of centers Collaboration Timeline Accrual goal

Are there competing protocols?

CALGB 100104 CALGB 100104 Revlimid After AutotransplantRevlimid After Autotransplant

Monthly AccrualMonthly Accrual

0

5

10

15

20

25

30

Apr-

05

Jun-0

5

Aug-0

5

Oct-

05

Dec-0

5

Feb-0

6

Apr-

06

Jun-0

6

Aug-0

6

Oct-

06

Dec-0

6

Feb-0

7

Apr-

07

Jun-0

7

Aug-0

7

Oct-

07

Dec-0

7

Feb-0

8

Apr-

08

Jun-0

8

Aug-0

8

Oct-

08

Launched 18 months into accrual of BMT CTN 0102 (auto-auto vs auto-allo transplants) which targeted same patient population

CALGB 100104 CALGB 100104 Revlimid After AutotransplantRevlimid After Autotransplant

Monthly AccrualMonthly Accrual

0

5

10

15

20

25

30

Apr-

05

Jun-0

5

Aug-0

5

Oct-

05

Dec-0

5

Feb-0

6

Apr-

06

Jun-0

6

Aug-0

6

Oct-

06

Dec-0

6

Feb-0

7

Apr-

07

Jun-0

7

Aug-0

7

Oct-

07

Dec-0

7

Feb-0

8

Apr-

08

Jun-0

8

Aug-0

8

Oct-

08

Complete accrual 2009; results released early

Don’t design protocol to failDon’t design protocol to fail

Eligibility criteria Balance between maximizing accrual

and minimizing heterogeneity that confounds outcomes

Examine the impact of specific criteria

Easy enrollment

Unbiased population

EXAMPLE – INCLUSION CRITERIA% of patients

100 Day survival

Acute leukemia, CR1

21% 16%

Acute Leukemia, CR2

11% 21%

CML, CP1 11% 13%CML, AP 5% 22%MDS 1% 22%MPS 1% 22%NHL / HD, CR1 2% 17%Acute leukemia, Rel 1

9% 30%30%

NHL / HD, PIF 6% 33%33%NHL / HD, Rel 1 7% 30%30%NHL / HD, CR2 1% 30%30%NHL / HD, Rel 2 2% 33%33%NHL / HD, Other 4% 40%40%CLL 3% 35%35%Multiple Myeloma 8% 40%40%

OK

NOTOK

Biological assignment trial-sometimes Biological assignment trial-sometimes necessary for a successful trialnecessary for a successful trial

Randomized trial: A random method is used to assign the treatment group, All patients must be eligible to receive

either treatment Balances unknowns that influence outcome

Biological assignment trial: Treatment group is assigned by availability of therapy HLA-identical Sib vs. Auto Sib vs. alternative donor Related vs. unrelated Related vs. auto

Biological assignment – Biological assignment – WHY?WHY?

BIOLOGIC ASSIGNMENT: Why

For Phase III trial needing 200 patients over 2 years, 25% of all eligible patients must be enrolled

7,000

2,0005,000

AMLOther disease

8001200

HLA-ident sibOther donor

400

in ~200 centers

400

CR1Not CR1

BMT CTN 0102: BMT CTN 0102: Auto-auto vs auto-allo HCT for MyelomaAuto-auto vs auto-allo HCT for Myeloma

NEED a Sib donorNEED a Sib donor% With Donor

Trial DesignAnnual Accrual

Sample Size Required

Years of Accrual

20%Randomized trial

(Matched sibling donors only)

51 462 9.1

Biological Assignment (all patients) 254 720 2.8

25%Randomized trial


63 462 7.3


30%Randomized trial


76 462 6.1


BIOLOGIC ASSIGNMENT – BIOLOGIC ASSIGNMENT – WHY NOT?WHY NOT?

Donor availability may influence outcome eg., very young patients may not have siblings eg., Older patients’ may not have healthy

available siblings Planned assessment and adjustment for major

prognostic factors Unblinded study: Physician or patient knowledge

of donor availability may differentially affect enrollment Influence referral if donor availability known

early Subtle pressure for or against enrollment Need to monitor for enrollment bias

Biologic Assignment: 0102Biologic Assignment: 0102

Number of Living SiblingsP value

0-1 ≥2

Age 0.001

≤55 45% 59%

>55 55% 41%

Race 0.3

Caucasian

82% 78%

Other 18% 22%

AFTER YOU OPEN THE PROTOCOLAFTER YOU OPEN THE PROTOCOL

Cheerleading – an essential element Personal communication is powerful

Get others to champion your protocol Acknowledge that accrual is difficult and

needs focused attention Figure out how to provide help Balance praise, cajoling, enticements,

appeal to higher calling

Accrual Plan

Accrual Plan AssessmentAccrual Plan AssessmentTarget Accrual and Accrual Period---monitor steps &

progress

Number of Sites Number of Committed SitesOptimal Number of Sites

Potential High Value Sites

Solicit cooperative group or special collaborators

Seek Patient Advocacy support

Anticipated Accrual BarriersCompeting ProtocolsCompeting Treatment Preference/Bias

Recruitment Focus:Transplant centersReferring MDsPatients

Never forgetAvoid the most annoying CROs

Tend to press detail over quality

If they don’t like the trial, they won’t enrollWork on inclusiveness prior to openingAnticipate accrual barriers

Pay enough to cover the costsSlow accrual is more costly than more $$/patient

Don’t be complacent; Ongoing communication helps. Guilt helps too

AFTER YOU OPEN THE PROTOCOLAFTER YOU OPEN THE PROTOCOL

Pay attention Identify problems early and correct them

quickly

Trial success is collaborative Listen to centers—before & after trial opens

Limit amendments Only those needed for safety or accrual

Confirmatory trials are essential !!!

-Reduce likelihood of false +

-Verify original results

-Expands the number of observations, perhaps allowing subgroup effects to be seen in combined analysis (Meta-analysis)

BUT-- only one trial doesn’t teach us the truth

Improving Accrual to Clinical Trials Saving patients and $aving Time Daniel Weisdorf October 2011 Thanks to Mary Horowitz.

Documents

hct slide

uncommon slide

eligible patients

hct trials

us centers

heterogeneous patients

cr1not cr1 slide

definitive trials