Annual Meeting Proceedings Part 1 - American Society of Clinical ...
Annual Meeting Proceedings Part 1 - American Society of Clinical ...
Annual Meeting Proceedings Part 1 - American Society of Clinical ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
152s Developmental Therapeutics—<strong>Clinical</strong> Pharmacology and Immunotherapy<br />
2541 General Poster Session (Board #1D), Mon, 8:00 AM-12:00 PM<br />
Relationship <strong>of</strong> variability in tumor measurement and response assessment.<br />
Presenting Author: Binsheng Zhao, Department <strong>of</strong> Radiology, Columbia<br />
University Medical Center, New York, NY<br />
Background: RECIST is widely used to evaluate anti-cancer therapy efficacy.<br />
This study explored variability in reporting tumor change and<br />
response to therapy due to both target lesion selection and measurement.<br />
Methods: 2256 measurements were performed in CT scans <strong>of</strong> chest,<br />
abdomen and pelvis from 30 patients retrospectively taken from a multicenter<br />
Phase II/III colorectal clinical trial. Using RECIST, three radiologists<br />
interpreted baseline, 6-wk and 12-wk scans in the following manner: (1)<br />
Radiologists independently selected and measured target lesions, (2) one<br />
radiologist’s target lesions were re-measured by the other two and (3) one<br />
radiologist re-measured the same scans in the above manner at an interval<br />
<strong>of</strong> greater than a month to prevent memory recall. Measurement variability<br />
in total tumor burden (TTB) on relative changes at 6-wk and 12-wk from<br />
baseline was analyzed for inter- and intra-reader target lesion selection,<br />
inter- and intra-reader measurement using Bland-Altman method, and<br />
agreements on RECIST categorical responses were assessed by kappa<br />
coefficient. Results: When the same target lesions were used, inter- and<br />
intra-reader variability in TTB on relative changes at 6-wk and 12-wk from<br />
baseline were similar; all had 95% limits <strong>of</strong> agreement within (-15%,<br />
15%). Kappa coefficients for RECIST were 0.74 (6-wk) and 0.87 (12-wk)<br />
for inter-reader and 0.64 (6-wk) and 0.88 (12-wk) for intra-reader to report<br />
responses. When radiologists independently selected and measured target<br />
lesions, variability in relative changes was within (-17%, 16%) at 6-wk and<br />
(-24%, 23%) at 12-wk for inter-reader and (-16%, 16%) at 6-wk and<br />
(-14%, 18%) at 12-wk for intra-reader interpretations. Kappa coefficients<br />
were 0.66 (6-wk) and 0.75 (12-wk) for inter-reader and 0.63 (6-wk) and<br />
0.80 (12-wk) for intra-reader to report responses. Conclusions: Differences<br />
exist in measuring tumor change. The magnitude <strong>of</strong> change in categorical<br />
RECIST response is quantifiable. The largest differences are when radiologists<br />
independently select and measure target lesions, the smallest when<br />
one radiologist repeats measurements on identical target lesions. The<br />
variability may impact the reporting <strong>of</strong> categorical responses and trial<br />
results, especially in a single arm study.<br />
2543 General Poster Session (Board #1F), Mon, 8:00 AM-12:00 PM<br />
<strong>Clinical</strong> importance <strong>of</strong> including new and nontarget lesion assessment <strong>of</strong><br />
disease progression (PD) to predict overall survival (OS): Implications for<br />
randomized phase II study design. Presenting Author: William Leonard<br />
Mietlowski, Novartis Oncology, Florham Park, NJ<br />
Background: Fridlyand (2011) retrospectively compared PFS vs. change in<br />
tumor burden as a primary endpoint in phase II non-small cell lung cancer<br />
(NSCLC) trials to inform phase III decision making and found the use <strong>of</strong><br />
PFS was superior. Since the classic tumor burden model only uses<br />
measurements <strong>of</strong> target lesions, we investigated whether the model could<br />
be strengthened by incorporating new and non-target lesion progression.<br />
The ability to use a strong tumor burden model has the benefit <strong>of</strong> potentially<br />
earlier decision making and considerable timeline savings. Methods: We<br />
analyzed five phase III trials <strong>of</strong> combination chemotherapy � targeted<br />
therapies with an OS primary endpoint: 1st, 2nd line NSCLC (ATTRACT-1,<br />
-2), 1st, 2nd line colorectal carcinoma (CONFIRM-1, -2), and 2nd line<br />
ovarian cancer (EPO906A2303). We applied Cox’s proportional hazards<br />
model to OS using the covariates <strong>of</strong> baseline tumor burden, 1st tumor<br />
assessment percentage change from baseline, new lesions, and non-target<br />
PD. Results: See table. Conclusions: We show that predictive models for OS<br />
should consider new and non-target lesions for PD, as well as target lesion<br />
tumor burden, findings independently corroborated by Suzuki (2011). We<br />
propose a longitudinal rank-based randomized phase II design, ranking a<br />
patient’s risk <strong>of</strong> death, differentially weighting PDs by type and time <strong>of</strong> PD,<br />
and percentage change in tumor burden. This may be more informative for<br />
phase II decision making for phase III trials based on OS, than PFS which<br />
only uses time <strong>of</strong> PD. Further studies with other tumor types and treatment<br />
modalities are warranted.<br />
Study<br />
Baseline<br />
tumor burden*<br />
Estimated hazard ratio (Wald test p value)<br />
% change<br />
from baseline* New lesion<br />
Non-target<br />
lesion PD<br />
ATTRACT-1 1.36 (p�0.001) 1.37 (p�0.001) 3.37 (p�0.001) 1.95 (p�0.001)<br />
ATTRACT-2 1.21 (p�0.001) 1.49 (p�0.001) 3.91 (p�0.001) 1.19 (p�0.332)<br />
EPO906A2303 1.09 (p�0.053) 1.12 (p�0.046) 2.20 (p�0.001) 1.67 (p�0.001)<br />
CONFIRM-1 1.31 (p�0.001) 1.38 (p�0.001) 3.02 (p�0.001) 1.40 (p�0.087)<br />
CONFIRM-2 1.41 (p�0. 001) 1.47 (p�0.001) 2.54 (p�0.001) 1.81 (p�0.001)<br />
* Unit <strong>of</strong> measurement is one standard deviation.<br />
2542 General Poster Session (Board #1E), Mon, 8:00 AM-12:00 PM<br />
A multicenter phase II neoadjuvant trial <strong>of</strong> bevacizumab combined with<br />
docetaxel plus carboplatin in the treatment <strong>of</strong> triple-negative breast cancer:<br />
Korean Cancer Study Group (KCSG-BR 0905, NCT 01208480). Presenting<br />
Author: Hye Ryun Kim, Division <strong>of</strong> Medical Oncology, Department <strong>of</strong><br />
Internal Medicine, Yonsei University College <strong>of</strong> Medicine, Seoul, South<br />
Korea<br />
Background: Triple negative breast cancer (TNBC) is dismal disease that is<br />
ineligible for endocrine or anti-HER2 therapy. We have conducted a phase<br />
II neoadjuvant trial to evaluate efficacy and tolerability <strong>of</strong> bevacizumab<br />
combined with docetaxel plus carboplatin in TNBC. Methods: Women with a<br />
histologically confirmed stage II or III TNBC were eligible. Hormone<br />
receptors and HER2 negativity were confirmed by immunohistochemistry<br />
and/or fluorescence in situ hybridization. Patients received 6 cycles <strong>of</strong><br />
docetaxel 75 mg/m2 , carboplatin AUC 5, and bevacizumab 15 mg/kg on<br />
day1 every 21 days. Bevacizumab was omitted in the last cycle to avoid<br />
wound complication. The primary endpoint was pathologic complete<br />
response (pCR) rate in both breast and axillary lymph node and secondary<br />
endpoints were clinical response rate, toxicity pr<strong>of</strong>iles, and breast conserving<br />
surgery (BCS) rate. Results: Forty-five TNBC patients were recruited<br />
from 7 institutes in the Korean Cancer Study Group (KCSG) between<br />
October 2010 and August 2011. The median age <strong>of</strong> the patients was 45<br />
years (30-72 years). The T1, T2, and T3 were 4.4 %, 68.9 %, and 26.7 %,<br />
respectively and axillary lymph node was positive in 80% by breast MRI.<br />
Among 45 patients, 44 patients completed 6 cycles <strong>of</strong> therapy followed by<br />
surgery and one patient was withdrawn due to patient’s refusal <strong>of</strong> further<br />
therapy after 3rd cycle. The pCR rate was 42.2 % (19/45) and clinical<br />
response rate was 95.5 % (43/45) (CR, n� 6; PR, n�37; SD, n�1; not<br />
evaluable, n�1) based on RECIST criteria 1.1. BCS was undertaken in<br />
77.7 % (35/45). The grade 3 or 4 adverse events were neutropenia (38),<br />
febrile neutropenia (4), vomiting (3), nausea (2), anemia (1), thrombocytopenia<br />
(1), stomatitis (1), gastrointestinal bleeding (1), and increased ALT<br />
(1). Only one patient experienced delayed wound healing after breast<br />
surgery. Conclusions: Bevacizumab in combination with docetaxel and<br />
carboplatin as neoadjuvant treatment provided an encouraging pCR rate<br />
(42.2 %) and a negligible wound healing problem after surgery. The<br />
adverse events were manageable.<br />
2544 General Poster Session (Board #1G), Mon, 8:00 AM-12:00 PM<br />
How to run survival-endpoint phase II trials with 25% to 40% fewer<br />
patients. Presenting Author: Michael J. Schell, H. Lee M<strong>of</strong>fitt Cancer<br />
Center & Research Institute, Tampa, FL<br />
Background: New molecular information, whether arising from expression<br />
pr<strong>of</strong>iling, identification <strong>of</strong> key mutations, or proteomic analyses is making<br />
the personalization <strong>of</strong> therapy increasingly feasible. This leads to the<br />
fragmentation <strong>of</strong> common disease entities into multiple molecular subtypes.<br />
Having sufficient numbers <strong>of</strong> patients to demonstrate the effectiveness<br />
<strong>of</strong> novel targeted therapies for each subtype has therefore become<br />
challenging, essentially impeding progress. Methods: Use <strong>of</strong> sample size<br />
determination s<strong>of</strong>tware allows for comparison <strong>of</strong> phase II trials with a target<br />
survival proportion at a given time T based upon Kaplan-Meier versus<br />
exponential survival methods. Results: The use <strong>of</strong> exponential survival<br />
fitting methods, in lieu <strong>of</strong> the currently implemented trial designs, to single<br />
arm phase II studies can routinely reduce patient numbers by 25-40%<br />
(Table) without compromising the statistical power. The biggest advantage<br />
is that survival information both before and after time T reduces the<br />
variability for the exponential fit estimate. This approach will lead to the<br />
saving <strong>of</strong> financial and patient resources for researchers. Moreover, the<br />
time saved will accelerate the approval <strong>of</strong> agents making the proposed trial<br />
designs ethically attractive. However, this approach is only applicable if the<br />
exponential distribution assumption behind the approach is valid.<br />
Conclusions: The exponential survival fitting method should replace the<br />
current standard approach in all instances unless the exponential distribution<br />
assumption is known to be false. Detailed examples and analysis will<br />
be presented.<br />
Percent savings obtained by using the exponential fit approach.<br />
(T,T*)<br />
HR (1,1) (2,2) (3,3) (4,4) (1,1.5) (1,2) (2,3) (2,4)<br />
2.0 5 15 25 36 30 41 31 38<br />
1.83 8 15 26 40 31 42 30 36<br />
1.67 9 12 28 38 32 42 28 33<br />
1.50 6 16 23 38 30 40 30 36<br />
Avg. 7 14 26 38 31 41 30 36<br />
T is scaled so T�1 is the median survival time; T* is the time through which the<br />
exponential fit is used; HR is the ratio <strong>of</strong> the alternative to null hypothesis<br />
median survival times.<br />
Visit abstract.asco.org and search by abstract for the full list <strong>of</strong> abstract authors and their disclosure information.