David cocker feasibility_and_web_mining

Data Mining to Make Global

Feasibility Assessment More

Reliable

David J. Cocker,

Senior Partner

MDCPartners, Belgium

Feasibility means different things to different people

This presentation

• Evolving clinical trial landscape

• Information newly available via the internet

• Public data sources to enhance feasibility

reliability.

Data Mining Disclosure

More data

• Can we leverage these expanding public

data sources?

• To fix these poor assumptions

Leverage information

INSC

TRRBE

C

JapicC

TI

DRK

S

REPE

C

Number trials

Pubs Ratio

Five year lag

Evolution of trial registry and publication ratio

Normalized publication

count

Trial start Pubs Registered

An avalanche of

new information

will descend upon

us

Feasibility on Feasibility

However, with a relatively sophisticated industry approach to

knowledge management, metrics and analysis…

Why do we get this so wrong, so often?

Trends show a moving academic and industrial

landscape

Classic problem but there is a classic solution

Time

Expenditure

Recru

itm

ent

Planned

Invest in in-depth feasibility

Delay

Opportunity cost

Over-run

%

%

Y1 Y2

Throwing more money at feasibility.

Will it improve reliability?

Problem

1

Problem

2

Bad assumptions still plague Pharma

Internal Clinical team assumptions

10 subjects per site

4 subjects per site

Scanned 750 trials,

60,000 patient mass

Two year delay

Company added another 67 sites

76 sites to recruit 750 patients

Need 188 sites to recruit 750 patients

Meta-analysis outcomes

A study in diffuse large B cell lymphoma subjects who recently completed R-CHOP therapy.

The simplest meta-analysis of a trial

registry would have mitigated this poor

initial assumption.

Applying meta-analysis to classic questions

Protocol

Patients with

the diseaseWhere do they live?

Number required

Selection of site

Selection criteriaAccess

ExperienceEquipment

Go

Country selection

Sites in area

which may be

suitable

The Environmental Trial Conveyor Belt

Experience

Equipment

Feasibility My trial is rolling

New

Studies

Publication

pre-emptionRetention

Regulatory

Drug

Supply

The practice

We cannot escape a rolling feasibility process

Hard points

• Number of eligible patients expected to recruit

• Concurrent trial workload, particularly at recruitment stage

• Previous experience in similar clinical studies

• Recruitment & retention in prior clinical trials

• Site personnel study experience and training

• Trial-required facilities such as laboratories and pharmacies

Adding a new component to the feasibility formula

Private historical data

Enrolment history

Start-up dynamics

Country performance

In-house predictive modeling tools

Predictions

Estimations

Meta-

Evidence

Internal KPI

History

Predictive modeling and

decision support tools

Survey data solicited from

potential sites

Best

Guess

Global transparency

Global trial activity

Academic literature

Disclosure

What’s out on the net and what’s to come?

• Regulatory push, societal expectation

– Sunshine Act and payments to healthcare professionals

– Clinical trial registries and result synopses

– Journal editors requiring registration

– Institutional review committees and procedures

Conclusion

More disclosure, more transparency, more to come!

Data Relationship and Semantics

Semantics System

Clinical

trial

Registry

FDA, EUHospital

Directory Commercial

Web portal

Pharmaceutical

company

Ad hoc

Web

Information

Conference

seminar

Chaos

Published

Investigator

Medline

OrderWorld demographics

+

Male Female

It’s not just about clinical research disclosure. It’s about the reality of internet

information linkied up.

Identifying experienced individuals in organizations

MeSH

Therapy relevance

Impact

factor

60 1 302 30

InvestigatorSiteSponsor

Epidemiology

ConditionDrugs Trials Treatment

use

Key data elements of the The power of semantic web disambiguation

Investigator

Site

Sponsor

Condition

Drugs

Trials

Treatment

use

A better view of the environment without the emotion

Investigator

Subject

Travelling

Distance(134

Km)

Incidence (189,000)

Screening

Failure (16746)

Female (189,000)

Age (167,456)

Popu

lation p

ool availa

bili

ty

Site load for area 770/

55 sites

Subject enrollment target 700

Breast Cancer Phase ll

Population Pool (210,000,000)

Classify system to research questions

Who

What

When

Where

Investigator

Site

Sponsor

ConditionDrugs

Trials

Treatment

use

Information that is on the move, stays on the

move. Monitor and re-visit often.

Number of investigators - 220

Number of investigators - 96

Regional population – 3,500,000

Essen as a region

Regional population – 7,500,000

Berlin

Investigator (score)

Investigator (score)

Trial Count (score)

Trial Count (score)

Investigator (score)Let the robot do the

legwork, and then debate

the assumptions.

Visualization of clinical trial registries

25

Disambiguating a trial registry can

render a nice picture

Rituximab sites

Breast Cancer sites

United Kingdom

Germany

Belgium

France

Traffic light system to

indicate site

availability

Site location

Estimated

enrollment

histogram

Organization

score based on

internet

footprint

Trial

experience in

years

Average

patients per site

Absolute number of

patients per site

accounting for

incidence, catchment

radius and screening

failure

Competing sites in

catchment area

based on site

criteria

Can you answer Questions

Ranking data, even if qualitative, allows a better

basis for discussion than a crystal ball.

Navigating complex interdependencies

Medical need The model is under stress

Better communicationMore trust

Conclusions

• An automated and rolling corporate engagement in site evaluation

and ranking.

• Mash-up and visualize all available data not just your own.

• Exploit expanding disclosure data as a tangible return on investment

for your participation.

• Validate your historic data with more dynamic data.

• Confirm assumptions through more targeted sampling based on

internet meta-analysis.

• Expand cross industry KPIs.

David J. Cocker

Senior Partner

Product Specialist Clinical Business Intelligence

Systems

MDCPartners cvba

Vluchtenburgstraat 5 2630 Aartselaar – Belgium

Office +32 (0) 3 870 97 50

Direct +32 (0) 3 870 97 72

Fax +32(0) 3 870 97 51

www.mdcpartners.be

Product www.ta-scan.com

Thank you