Upload
rp
View
213
Download
0
Embed Size (px)
Citation preview
Volume 90 � Number 1S � Supplement 2014 Oral Scientific Sessions S9
voxel size of the planning CT. The dose volume was then scaled to reflect
dose delivered after 1 week, 2 weeks, and 3 weeks. The region of interest
(ROI) was defined as the bony pelvis shrunk by a uniform 5 mm margin.
The reduction in ROI volume removed a region that was most likely
cortical bone and reduced potential noise due to partial volume effect from
the large voxel size of the FLT PET image. The resampled FLT images
were used to calculate the voxel by voxel value change in FLT activity as a
surrogate for marrow activity which was then correlated to dose in 1 Gy
increments. The results for each subject were plotted on the same frame of
reference for inter subject comparison.
Results: Voxel by voxel analysis of bone marrow activity change repre-
sented in time series FLT PET images shows an exponential decrease
in FLT SUV during chemoradiation therapy when normalized by the
pre-therapy image for all subjects. This relationship can be fit by the
exponential equation FLTn Z 1.3 e-1.17(Dose), where FLTn Z weekly
normalized FLT SUVand DoseZ radiation dose (Gy) at the time of FLTn.
The average coefficient of determination (R2) Z 0.84 for all 21 subjects.
Out of 21 subjects, 19 had R2 > 0.7. This relationship appears to be
independent of time or fractionation. FLT SUV voxel values decreased
50% when compared to pretherapy values for a range of radiation doses
between 2.2 Gy - 6.8 Gy for all subjects after 1, 2, or 3 weeks. The mean
dose to reach 50% of pretherapy FLT SUV was 4.5 Gy. However, the range
in FLT SUV voxel change decrease after receiving 4.5 Gy was 25 - 83%
for all subjects after 1, 2, or 3 weeks suggesting significant patient vari-
ation that was not related to fraction size.
Conclusions: Bone marrow activity changes as a function of dose can be
measured using voxel by voxel analysis of time serial FLT PET images.
This method allows quantitative analysis of FLT PET images at a reso-
lution equal to the voxel size in the PET image for determining the rela-
tionship between FLT uptake change and radiation dose.
Author Disclosure: S.M. McGuire: None. R. Hareendran: None. J. Xia:
None. S. Bhatia: None. W. Sun: None. W.M. Rockey: None. Y. Menda:
None. L. Ponto: None. B. Gross: None. J. Buatti: None.
9Initial Scale of MRI Heterogeneity of Pediatric Rhabdomyosarcomais Prognostic for RecurrenceM.F. Gensheimer,1 A.D. Trister,1 D.S. Hawkins,2 and R.P. Ermoian1;1University of Washington Medical Center, Seattle, WA, 2Seattle Children’s
Hospital, Seattle, WA
Purpose/Objective(s): In many cancers, intratumoral heterogeneity has
been found in histology, genetic variation, and vascular structure. We
developed a novel algorithm to interrogate different scales of heterogeneity
using standard clinical MR imaging. We hypothesize that the physical
scale of this heterogeneity holds important clues to disease aggressiveness.
Specifically, heterogeneity at large distance scales may correlate with
treatment resistance and propensity for disease recurrence. We applied our
algorithm to initial diagnosis MRI of rhabdomyosarcoma patients for
prediction of recurrence.
Materials/Methods: The Spatial Heterogeneity Analysis by Recursive
Partitioning (SHARP) algorithm recursively segments the tumor image
into increasingly smaller regions. The tumor is repeatedly subdivided, with
the dividing line chosen so as to maximize signal intensity difference
between the two regions. This process continues until the tumor has been
divided into single voxels, resulting in segments at multiple voxel scales.
For each scale, heterogeneity is measured by comparing each segmented
region to the adjacent region and calculating the difference in signal in-
tensity histograms. In an IRB-approved retrospective analysis, we
measured the scales of heterogeneity of primary tumor of 18 rhabdo-
myosarcoma patients on initial diagnosis MRI. Using univariate Cox
proportional hazards regression, we explored the influence of heteroge-
neity parameters on relapse-free survival (RFS). Patients were treated
using national standard regimens.
Results: Most patients had COG intermediate-risk disease (n Z 10). Most
common disease sites were parameningeal (nZ 6) and orbit (nZ 5). With
35 month median follow-up, there were 10 disease recurrences. On T1-
weighted gadolinium-enhanced MRI, larger scale of maximum signal in-
tensity heterogeneity, relative to maximum tumor diameter, was prognostic
for shorter RFS (Cox model log-rank p Z 0.05). Patients with scale of
maximum heterogeneity greater than median (n Z 7) had median RFS of
15.5 months, versus not reached when scale of maximum heterogeneity
was less than or equal to median (n Z 11), log-rank p Z 0.02. Slope of
heterogeneity versus scale was not prognostic for RFS (Cox model p Z0.45). Clinical covariates, including histology and COG risk group, were
also not prognostic.
Conclusions: The SHARP algorithm produces a biologically motivated
segmentation of tumor regions and reports the amount of heterogeneity at
various distance scales. In rhabdomyosarcoma, RFS was shorter in patients
with primary tumors exhibiting larger scale of heterogeneity on contrast-
enhanced MRI. If validated on a larger dataset, this imaging biomarker
could be useful for prognostication and to help personalize treatment.
Author Disclosure: M.F. Gensheimer: None. A.D. Trister: None. D.S.
Hawkins: None. R.P. Ermoian: None.
10Serum Vascular Endothelial Growth Factor-A and TransformingGrowth Factor-b1 Can Predict Pathological Response and Disease-Free Survival of Esophageal Cancer Patients Treated withNeoadjuvant Chemoradiation Therapy Followed by EsophagectomyY. Chiang,1 J. Cheng,1,2 M. Graber,3 F. Hsu,1 C. Tsai,1 J. Lee,4 D. Chang,3
and A. Koong3; 1Division of Radiation Oncology, Department of
Oncology, National Taiwan University Hospital, Taipei, Taiwan,2Graduate Institute of Oncology, National Taiwan University College of
Medicine, Taipei, Taiwan, 3Stanford University, Stanford, CA, CA,4Department of Surgery, National Taiwan University Hospital, Taipei,
Taiwan
Purpose/Objective(s): This study is aimed to identify serum biomarkers
that predict treatment response and survival by screening proximity liga-
tion assay (PLA) and verified enzyme-linked immunosorbent assay
(ELISA) for patients with esophageal squamous cell carcinoma (ESCC)
undergoing neoadjuvant concurrent chemoradiation therapy (CCRT) fol-
lowed by esophagectomy.
Materials/Methods: One hundred three patients with ESCC receiving
CCRT consisting of taxane-/5-fluorouracil-based chemotherapy and 40 Gy
radiation therapy followed by surgery were prospectively enrolled. Serum
samples were collected before and within 1 month after completion of
CCRT. With the use of PLA, 15 biomarkers were simultaneously analyzed
in the initial 79 patients. The biomarkers significantly associated with
pathological response (PathR)/survival were verified by ELISA in an
expanded group of 103 patients. Associations between serum levels of
biomarkers and clinical factors correlating with PathR, disease-free sur-
vival (DFS), and overall survival (OS) were evaluated by ANOVA and log-
rank tests.
Results: Following CCRT, 38 patients had pathologically complete
response (37%), 44 microscopic (43%), and 21 macroscopic residual
disease (20%). With a median follow-up of 33.7 months, the median DFS
and OS were 21.9 months and 42.3 months, respectively. Among the 15
biomarkers screened by PLA, vascular endothelial growth factor-A
(VEGF-A) and transforming growth factor-b1 (TGF-b1) were significantlyassociated with PathR and/or DFS. These biomarkers were further
analyzed by ELISA to confirm initial biomarker findings by PLA. On
ELISA, both pre- and post-CCRT VEGF-A levels were significantly
correlated with PathR (p Z 0.042 and 0.019, respectively). Patients with
pre-treatment VEGF-A less than 250 pg/ml were more likely to have
pathologically complete response after CCRT than VEGF-A more than
250 pg/ml (20/35 vs. 18/68, p Z 0.002). Patients with high pre-CCRT
VEGF-A/TGF-b1 levels (� median) had significantly worse median DFS
(9.7 months vs. 42.9 months, p Z 0.009) and worse median OS (19.2
months vs. 46.2 months, p Z 0.07). On multivariate analysis, PathR (p Z0.003) and pre-CCRT high levels (� median) of both VEGF-A and TGF-