Statistics in Pharmaceutical Development

1
Statistics in Pharmaceutical Development

• Buffy Hudson-Curtis
• Statistical Sciences, RTP
• GlaxoSmithKline
• 12-September-2005

2
(No Transcript)
3
(No Transcript)
4
(No Transcript)
5
(No Transcript)
6
Statistical Sciences RTP

• Support scientists in pre-clinical and
  non-clinical development.
• Provide statistical consulting and training.
• 8 Full-time Statisticians
• 3 Part-time Graduate Industrial Trainees
• Similar groups in Philadelphia and the UK
• Specialized each statistician supports one
  area.

7
Our clients

• Clients include B.S., M.S., and Ph.D Chemists,
  Engineers, Biologists and Pharmacists.
• Using statistics is usually not mandated in
  pre-clinical and non-clinical work.

8
Role of non-clinical statistician

• Act as an internal consultant to areas within
  drug development.
• Large number and variety of small problems
• Wear many different hats statistician,
  programmer, teacher, presenter, leader, planner,
  and negotiator, to name a few.
• Must quickly provide easy-to-understand and
  easy-to-apply high quality solutions.
• Must be able to think on your feet.

9
Typical Projects

• DOE
• Analysis of data
• Comparisons (t-tests, equivalence tests)
• Setting specifications (tolerance intervals,
  prediction intervals, probability calculations)
• Instruction (2-day courses, lunch-and-learn)
• Training in statistical software (Excel, JMP,
  Statistica)

10
Allocation of Time

• 20 meeting with scientists
• 40 statistics
• 30 writing reports/presentations
• 10 administrative

11
Typical Project in Pharmaceutical Development

• Design Experiment
• 1-2 hours with scientist
• Meet with scientist, understand problem (goal,
  constraints), visit lab, see equipment.
• Challenge resources are limited. Fewest number
  of runs to get the most information.
• May need to perform sequential experimentation.
• Possible designs include D-optimal, fractional
  factorials, full factorial, central composite

12
Typical Project in Pharmaceutical Development

• Analyze Data
• 1-8 hours
• Discuss with scientist
• Many responses
• Optimize competing responses
• Maximize yield, minimize impurities
• Communicate Results to Scientist
• 2-8 hours
• Write report
• Emphasis on graphics and practical results (not
  p-values)

13
Typical Project in Pharmaceutical Development

• Follow up with Scientist
• Discuss report
• Plan next steps
• Future Experiments
• Confirmation Runs
• Power (or lack of)
• Lurking variables
• Applications outside of design space

14
Useful courses and Experiences at NCSU

• Design of Experiments!!!!!!
• Inference
• Non-parametric, linear models
• Take advantage of opportunities to teach or
  present
• Graduate Industrial Trainee

15
(No Transcript)