![](3D"slide0001_image001.gif")
<=
/span>
=
Inter-observer
Variability in Radiographic Assessment of Response in Chemotherapy Trials =
for
Pancreatic Cancer
Brian C. Baumann, AB<= span style=3D'font-family:Arial;mso-ascii-font-family:Arial;mso-bidi-font-famil= y: "Times New Roman";mso-hansi-font-family:Arial;font-size:111%;color:white; mso-color-index:0;position:relative;top:-.45em;mso-text-raise:30%'>1, Robert R. Ford, MD1, Kathy Dasse, PharmD = 2, Kathy Zhou, PhD 3, Ruth Feldman, MD1 <= /span>
Brian C. Baumann, AB<= span style=3D'font-family:Arial;mso-ascii-font-family:Arial;mso-bidi-font-famil= y: "Times New Roman";mso-hansi-font-family:Arial;font-size:111%;color:white; mso-color-index:0;position:relative;top:-.45em;mso-text-raise:30%'>1, Robert R. Ford, MD1, Kathy Dasse, PharmD = 2, Kathy Zhou, PhD 3, Ruth Feldman, MD1 <= /span>
1 RadPharm Imaging Core Lab (Princet=
on,
NJ), 2 Pfizer, I=
nc.
(Ann Arbor, MI), 3 Department of Public Health, Weill Cornell Medical College (New York, N=
Y)
Discussion &am=
p;
Conclusions
![](3D"slide0001_image002.gif")
Discussion
•71% decrease in response ra=
te
in this study was even more pronounced than previously <=
span
style=3D'position:absolute;top:75.45%;left:67.5%;width:32.49%;height:1.8%'=
>reported =
23
– 52% decreases in response rates in studies of other primary
malignancies
•Possible explanations for the significant discrep=
ancy
between investigators and independent reviewers:
<=
/span>
•Investigator bias: Limited
alternative treatment options for unresectable pancreas cancer =
may bias
investigators to assess a more favorable response that would ensure patien=
ts continued
participation in the trial
=
•Challenge of accurately
measuring pancreatic cancer lesions with existing CT-based imaging
technology
Conclusions
<=
/span>
=
•This study supports the use of independent review=
of
trial data when the endpoint of the trial, =
e.g. response rate, is based=
on
imaging data
<=
span
style=3D'font-family:Arial;font-size:108%;visibility:hidden'>•<=
/span>
<=
/span>
Materials
and Methods
•133 patients in two
multi-center, randomized Phase II drug trials had their response assessed =
by
measurement of tumor size or new lesions on serial CT scans
•Protocols specified techniques of tumor measureme=
nt
& defined responses as complete (CR) or partial respo=
nses
(PR), stable disease (SD) or progressive disease (PD)
•Objective responses were confirmed by repeat imag=
ing
evaluations after an interval >4 weeks
=
•Early withdrawal of PR or SD cases from the study
<8 weeks after enrollment were scored as non-evaluab=
le
(NE)
•The same images were then blindly reviewed by
protocol-trained radiologists at RadPharm, an independent imaging core lab in
Princeton, NJ
•Cases of discordance in the interpretation of best
response between investigators and reviewers were then
blindly re-reviewed by an independent adjudicator
<=
span
style=3D'font-family:Arial;font-size:117%;visibility:hidden'>•<=
/span>
<=
/span>
NE
PD
SD
PR
CR
<=
span
style=3D'font-size:16%;mso-special-format:lastCR;display:none'>
15
45
67
6
0
Revie=
wers
18
44
50
20
1
Inves=
tigators
![](3D"slide0001_image004.gif")
Results
•In 50/133 cases, there was
sufficient discrepancy between investigators and reviewers=
to
result in a change in overall response
•In 10 of these cases, investigators reported
responses inconsistent with protocol standards for the investigator’s
reported tumor measurements
•In 40 cases, reviewers disagreed with the
radiological findings reported by investigators
<=
span
style=3D'font-family:Arial;font-size:117%;visibility:hidden'>•
<=
/span>
Best
Response Corrected to Conform to Protocol Criteria
•Discrepancies between
investigators and reviewers could be classified into the
following categories:
<=
span
style=3D'font-family:Arial;font-size:117%'>•Improperly
measured target lesions
<=
span
style=3D'font-family:Arial;font-size:117%'>•Incorrect
selection of target lesions
<=
span
style=3D'font-family:Arial;font-size:117%'>•Failure
to identify new lesions
<=
span
style=3D'font-family:Arial;font-size:117%'>•Missing
data
•Incorrect
response form paradigm (disease identified at baseline not subsequently followed, yet response was
reported)
<=
span
style=3D'font-family:Arial;font-size:117%'>•Failure
to apply the protocol standards correctly
<=
span
style=3D'font-family:Arial;font-size:117%;visibility:hidden'>•<=
/span>
<=
span
style=3D'font-family:Arial;font-size:117%;visibility:hidden'>•<=
/span>
<=
/span>
Background Information
<=
span
style=3D'font-family:Arial;font-size:117%'>•In
the era of evidence-based medicine, the quality of the evidence is
critical
•Previous
studies examining inter-observer variability in the radiographic assessmen=
t of drug response revealed significant
discrepancies
•Response
rates reported by investigators (INV) were significantly higher than rates=
reported by an independent review committee (IRC)
•No
studies have examined differences between investigators and independent
central =
review
panels for the imaging assessment of drug response in unresectable pancrea=
tic cancer
<=
span
style=3D'font-size:117%;visibility:hidden'>•
Summary
Table – Previous Studies
•In 47 of the 50 cases of
discrepancy between investigators and reviewers, the films were
available for a second review by the adjudicator
•The adjudicator confirmed the reviewers’
assessment of response in 39/47 (83%) of cases
•Discrepancies in response were not randomly
distributed between investigators & reviewers
•Investigators reported significantly more objecti=
ve
responses (CR + PR) than reviewers even after investigators’ response
coding errors were corrected
=
•Investigators reported favorable responses in 21/=
133
(16%) of cases compared with 6/133 (5%) responses noted by reviewers, a
significant discrepancy (p=3D.01)
<=
span
style=3D'font-family:Arial;font-size:117%;visibility:hidden'>•
<=
/span>
![](3D"slide0001_image006.jpg")
![](3D"slide0001_image008.jpg")
Examples of Discrepancies
Subject
1056 - Baseline
Subject
1056 – Baseline + 1
![](3D"slide0001_image010.jpg")
Subject
1021 – Baseline
Subject
1056
<=
span
style=3D'font-family:Arial'>• Investigator
Response: SD; Reviewer Response: PD; Adjudicator Response: PD
•Cause of
discrepancy: Investigator improperly measured the para-aortic nodal mass
target lesion & neglected other
target lesions, e.g. nodal mass to left of sma & portacaval node
<=
/b>
•At baseline, t=
he
investigator obtained a measurement of 63.0 sq cm by combining two adjacen=
t lesions. At baseli=
ne +
1, the investigator measured only one of the now clearly distinct lesions,=
artificially increasing the % drop in bidimensional lesion area
which contributed to the investigator’s =
incorrect response of SD.
=
Subject 1021
<=
span
style=3D'font-family:Arial'>• Investigator Response: SD; Reviewer Response: PD; Adjudicator Response: PD
•Cause of
discrepancy: Investigator improperly measured the pancreatic mass, includi=
ng
the dilated pancreatic duct
proximal to the mass in his measurement
•At baseline, t=
he
investigator measured the pancreatic mass as 33.6 sq cm compared to the
reviewer who measured 1=
7.5
sq cm. Whereas the reviewer =
noted
an increase in the size of the lesion after baseline, the investigator noted incorrectly that the mass was
decreasing.
<=
/span>
Discussion
& Conclusions