Poster - Vanderbilt University

advertisement
Poster Session: 3
Poster Number: 328
A Mediating Effect of AKT in EGFR Signal to Osteosarcomas
in a Clinical Study
Huiyun Wu, Adriana Gonzalez, and Yu Shyr
A National Cancer Institute-Designated Cancer Center
Vanderbilt-Ingram Cancer Center, Vanderbilt University, Nashville, TN 37232
ABSTRACT
A mediation model was created with structural
equation model (SEM) to analyze the function
pattern of EGFR and AKT signaling in a
osteosarcomas clinical study. The results
suggested a mediating effect for AKT by showing
significant associations for EGFR to Ki67
(p=0.0266), EGFR to AKT (p=0.0016), and AKT
to Ki67 controlling EGFR (p=0.0035). After the
impact of EGFR on Ki67 was carried by AKT, the
relation between EGFR and Ki67 was no longer
significant (p=0.4425). The mediating effect was
verified with Sobel test (p<0.001). The study
indicated that mediation model could be an
interesting procedure in testing a biologically
identified signal pathway with clinical data.
Structural Equation Model
RESULTS
Effects table
0, 1
e2
AKTpost
Age
EGFR → AKT
EGFR → Ki67
AKT → Ki67
0.850 (0.0016)
0.038 (0.0266)
0.034(0.0035)
Direct effects
0.850
0.009 (0.4425)
0.034
Indirect effects
0.000
0.029
0.000
Ki67post
0, 1
0, 1
d1
e1
Mediation model
A mediation effect occurs when the third variable (mediator) carries
the influence of a given independent variable (X) to a given dependent
variable (Y).
Concept and application were mainly developed in social science,
behavioral science, and preventive medicine.
Mediation models explain “how” an effect occurred by hypothesizing a
causal sequence.
Total effects
Data are expressed as coefficients and p values.
Total effect of EGFR → Ki67 =  x  + ’ =0.85 x 0.034 + 0.009 = 0.038
Indirect effects of EGFR → Ki67 = 0.850 x 0.034 = 0.029
Direct effects of EGFR → Ki67 = 0.009 which is the estimate after adjusting for AKT.
EGFRpost
INTRODUCTION
gender
Path
EGFR Signal Pathway
Statistical tests
Test
p value
Sobel
Goodman (I)
Goodman (II)
0.00000152
0.00000172
0.00000133
Causal sequence
Signaling events are ordered both spatially and temporally
The p values from the above 3 models indicated that criteria for
mediating effects were satisfied.
Statistical tests
Mediator
The three significant tests suggested a mediating effect for AKT.
X
Y
R R
pY
PI3-K
The intervention program (X) is designed to change mediating
variables (M) hypothesized to be causally related to the outcome (Y).
K KpY
RAS
SOS
pY
GRB2
STAT
AKT
PTEN
RAF
MEK
CONCLUSIONS
MAPK
Gene transcription
Cell cycle progression
PP
myc
cyclin D1
Cyclin D1
DNA
JunFos
Three models
• Statistical modeling could be another approach to elucidating a
mechanistic causal relation between events observed in biology.
Myc
Y = 0(1) +  X + (1)
(1)
Y = 0(2) + ’ X + M + (2)
(2)
M = 0(3) + X + (3)
(3)
Proliferation/
maturation
Survival
(anti-apoptosis)
Angiogenesis
Metastasis
STUDY DESIGN AND METHODS
Significance tests with products of coefficients of the models
69 patients of primary OS were reviewed.
54 samples were collected.
All patients completed one course of chemotherapy procedure.
a.
Sobel test
z-value =
b.
Goodman (I) test
z-value =
  22   2
Goodman (II) test
z-value =
  22   2
c.
Sources: MacKinnon & Dwyer (1994) and MacKinnon, Warsi, & Dwyer
(1995)
Initial variable EGFR and mediating variable AKT were
immunostaining index.
Outcome variable was cancer development expressed as Ki67 which
was also immunostaining index.
Statistical analysis was performed by creating SEM with AMOS.
• Mediation model might be a useful tool to identify the EGFR-Akt or
other pathway leading to cancer development.
Download