Electronic Data Acquisition in Clinical Trials

1
Electronic Data Acquisition in Clinical Trials

• Alan Young Chris Bray
• CTSU

2
Data Wanted!

• Better
• Cheaper

3
Workshop Plan

• Demonstration of technologies available
• Example scenarios
• Group presentations and discussion
• Comparisons
• Summary and round-up

4
Principal Distinction

• Proxy Methods user passes data to Centre in
  non-electronic format for conversion and storage
  by a member of Centre
• Direct Methods user enters data into Central
  computer systems without any other manual
  assistance

5
Principal Distinction

• Direct
• Web Browser
• Mobile Phone
• Laptop
• Palm Pilot

• Proxy
• Paper
• Telephone
• Optical Recognition

6
Comparison Factors

• Financial - can we afford to build, deploy and
  maintain a system
• Quality - how good is the data we will obtain
• Security - is it private and robust
• Experiential - is use sufficiently easy that
  people complete first entry and are willing to
  use it again

7
1. Paper

• User enters data onto a paper form
• Form is mailed/faxed to Centre
• Operator re-types data into Centre computer

8
Paper Characteristics

• Good
• Paper forms ubiquitous
• Cheapest set-up

• Bad
• No quality control at source
• Bulky to store
• Error prone to transcribe
• Easy to intercept or lose

9
2. Telephone

• User telephones Centre
• Telephonist interrogates user via structured QA
  dialogue

10
Telephony Centre
11
Trial Menu
12
Data Entry
13
Telephony Characteristics

• Pro
• Telephones ubiquitous
• Quick access
• Interaction with human operator gives reassurance

• Con
• Difficult for non-English speakers
• Staff must work on 24-hour basis for
  international studies
• Verbal transcription errors arise
• Lines can saturate

14
3. Optical Character Recognition

• User enters data onto a specially designed paper
  form
• Form is mailed/faxed to Centre
• Computer attempts to interpret and load form data

15
Optical form definition
16
Optical Data Checking
17
Optical Characteristics

• Pro
• Paper forms ubiquitous
• Much higher throughput than manual re-typing
• Electronic storage of raw forms facilitated

• Con
• Impossible to quality control at source
• Form design constrained
• Difficult to handle free-text
• Manual supervision required

18
4. Web Browser

• User connects via browser to secure web-page
• User enters data into special form
• Web server validates and stores data

19
Web Characteristics

• Pro
• Instant validation of data
• Users tend to self-train
• Bulk handling easy
• Data instantly available
• No paper to be stored

• Con
• Requires internet connection
• Authentication distribution

20
5. Mobile Phone (WAP)

• User dials secure WAP-page
• Use enters data into special form
• WAP server validates and stores data

21
Mobile Phone Characteristics

• Pro
• Instant validation of data
• Bulk handling easy
• Convenient to carry
• Data instantly available
• No paper to be stored

• Con
• Requires WAP connection
• Authentication distribution
• Clumsy interface

22
6. Laptop

• Patient interviewed in-situ by trained operative
• Data entered directly into laptop
• Data transferred electronically to CTSU

23
Laptop Characteristics

• Con
• Expensive equipment
• Complicated development cycle

• Pro
• Instant validation of data
• Data rapidly available at Centre
• No paper to be stored
• Available off-line

24
7. Palm Pilot (PDA)

• Patient is loaned Palm
• Patient enters data into Palm
• Data transferred electronically to CTSU

25
Palm enter meal
26
Palm enter glucose
27
Palm daily record
28
Palm Characteristics

• Con
• Expensive equipment
• Complicated development cycle
• Easy to lose equipment
• Variety of interfaces in use

• Pro
• Instant validation of data
• Data rapidly available at CTSU
• No paper to be stored

29
Example Scenarios

• There are 4 scenarios
• Your group has been assigned 2 of the scenarios
  to study and discuss
• Decide which technologies you feel would be most
  appropriate for these
• Present your conclusions to the other groups

30
Financial Factors

• Set-up - how expensive is the system to develop
  in terms of programming, equipment and staff
  training
• Running - how expensive is the system to run on
  an ongoing per-unit basis

31
Quality Factors

• Accuracy - how well does it record what the user
  intends it to
• Consistency - how good is it at spotting and
  correcting logical inconsistencies in data
  entered
• Accessibility - what fraction of potential users
  can make use of a method

32
Security Factors

• Robustness - how well does it survive equipment
  failures
• Privacy how easy is it for unauthorised persons
  to view or alter the data
• Regulatory what external constraints have to be
  satisfied

33
Experiential Factors

• Small datasets - is it good for entering things
  like the time of an appointment
• Large datasets - is it good for entering things
  like a complete medical history
• Feedback - can it give users feedback on their
  input, such as a randomization allocation
• Pressure - can users enter data at a time and
  rate of their own choosing

34
Comparison Summary
Higher values are better
35
Summary

• No method is optimal for a majority of tasks
• Target audience should be considered when
  selecting methods
• Many projects will benefit from multiple entry
  channels
• Particular factors may over-ride all other
  considerations

36
Conclusion

• Treat computing as an integral part of conducting
  a large study, not an add-on extra.
• Consider methods of data collection at the very
  first stages of study design

37
Relevant regulations and guidelines

• ICH-GCP (1996)
• FDA regulation 21 CFR Part 11 regulation (1997)
• FDA guideline Computerized systems used in
  clinical trials (1999)

38
Requirements for electronic records

• Validation
• System access restrictions
• Audit trails
• Record review and record protection

39
Requirements for electronic signatures

• Can be biometric or non-biometric
• Signature must be linked to record
• Shall be unique to an individual
• Individual's identity must be verified
• FDA require "Certification Letter" prior to use

40
Non-biometric signatures

• Minimum of 2 components (ID/password)
• Only to be used by genuine owner
• Misuse requires collaboration of 2 or more