Jessica A. Behrle, Senior Principal Biostatistician at Janssen R&D Johnson & Johnson My
Jessica A. Behrle, Senior Principal Biostatistician
at Janssen R&DJohnson & Johnson
My
http://thisisstatistics.org/Why Study Statistics
My career path to Stats
LOVED MATH
Exposure to StatsResearch Experience for Undergraduates
Co-op
More Stats
What Do I Do?
• Statisticians are doing “detective work- numerical detective work - or counting detective work - or graphical detective work… “ by John Tukey (famous statistician)
What are my Duties?
• Determine the questions or problems to be addressed and the data needed to answer them
• Determine methods for finding or collecting data• Design ways to collect data • Analyze and interpret data• Report conclusions from their analyses
What Skills Are Important?
Quantitative
Problem Solving
Computer
Communication People
Nonclinical Statistics
• Support– Preclinical/Discovery– Manufacturing
• Consultant• Teach
Lead Target Safety Phase I Phase II Phase III Post Marketing
Preclinical Clinical
10
Rat’s Eye
Retina
Special Acknowledgements to the creator: Ray Heslip , SIDD
Number of Photoreceptors (i.e., Thickness) was measured at locations across
the retina.
11
Temporal to NasalT4 T3 T2 T1 N1 N2 N3 N4
Superior ToInferior
S4
S3 12 10 10 11 11 11 10 11
S2 10 12 17 24 14 10 10 6
S1 11 10 10 11 10 10 17 13
I1 11 10 10 10 10 19 9 11
I2 12 12 13 13 15 22 11 10
I3 10 20 11 11 11 12 10 13
I4
Ex. # Photoreceptor Cells by Quadrant (i.e., Thickness)
TEMPORAL/SUPERIOR
(TS)
NASAL/SUPERIOR
(NS)
TEMPORAL/INFERIOR
(TI)NASAL/INFERIOR
(NI)
12
Treatments
Group Name
N Strain of Rat
Surgery Location
Low dose 12 Dys Sub-retinal
Mid dose 12 Dys Sub-retinal
High dose 11* Dys Sub-retinal
WT 12 WT NA
Untreated 12 Dys NA
Sham 11* Dys Sub-retinal
Histology was obtained from 4 rats on day 30, 4 rats on day 60, and 4 rats on day 90.
*Histology was obtained from only 3 rats on day 30.
13
Stat Methods of 1st Analysis
· where― is Grand Mean― Trti is Treatment Effect; i=high, mid, low, Sham, Untreated, and Wild Type (WT)
― Dayj is Day Effect; j=30, 60, and 90
― Quadrantk is Quadrant Effect; k=TS, NS, TI, and NI
― Ratl is Random Effect for Rat; l=rat id#
― ijkl is residual error
· Pairwise comparisons of treatments
· =0.05
ijklkjikj
kijilkji
QuadrantDayTrtQuadrantDay
QuadrantTrtDayTrtRatQuadrantDayTrtThickness
log
· Average thickness computed per quadrant and then modeled
14
Observed Thicknesses
15
Results of 1st Analysis
· High dose ↑ Untreated― 78.2 ± 12.5%
· Mid dose ↑ Untreated― 73.9 ± 11.9%
· Low dose ↑ Untreated― 57.3 ± 10.7%
· No overall significant difference is detected between the 3 doses.
― Largest diff High vs Low: 13.3 ± 7.9% (p-value=0.081)
· Each dose has significant ↓ compared to WT.
Overall
16
Quadrant Results: Doses vs. Untreated
· Day 90 · Day 60
TS ↑
TI ↑ NI ↑smallest
NS ↑
Arrows indicate statistically significant difference
TS ↑
TI ↑
NS
NI
Some of Statistical Analyses involved in Manufacturing
• Design - DOE• Comparability• Estimate Shelf-Life• Set Specifications
Lead Target Safety Phase I Phase II Phase III Post Marketing
Preclinical Clinical
Manufacture
Purification
http://www.buildingbiologics.com/manufacturing-biologics.html
Purification DOEsColor Map On Correlations
X1
X2
X3
X4
X5
X6
X1*X
2X1
*X3
X1*X
4X1
*X5
X1*X
6X2
*X3
X2*X
4X2
*X5
X2*X
6X3
*X4
X3*X
5X3
*X6
X4*X
5X4
*X6
X5*X
6
|r|0
1
|r|0
1
X1
X2
X3
X4
X5
X6
X1*X
2X1
*X3
X1*X
4X1
*X5
X1*X
6X2
*X3
X2*X
4X2
*X5
X2*X
6X3
*X4
X3*X
5X3
*X6
X4*X
5X4
*X6
X5*X
6
Difference between 2 sites
• Manufacturing a product in NA and EU• Question to address: Is there a difference in
the manufacturing process in these 2 regions?• Collect the data from the 2 regions.
– 20 measurements from each region– Measurement has values from 0 – 100 (↑ good)
• Combine, Understand and Show
JMP Data Table
• Consulting Courses• Save your text books (you may need them
later)• Learn how statistics can be used in fields that
interest you• Internships/Co-ops
ADVICE
My
NOW IT IS TIME TO DEVELOP YOUR: