Posts Tagged ‘tumor heterogeneity’

Live Notes, Real Time Conference Coverage 2020 AACR Virtual Meeting April 28, 2020 Session on Evaluating Cancer Genomics from Normal Tissues Through Metastatic Disease 3:50 PM

Reporter: Stephen J. Williams, PhD

 Minisymposium: Evaluating Cancer Genomics from Normal Tissues through Evolution to Metastatic Disease

Oncologic therapy shapes the fitness landscape of clonal hematopoiesis

April 28, 2020, 4:10 PM – 4:20 PM

Kelly L. Bolton, Ryan N. Ptashkin, Teng Gao, Lior Braunstein, Sean M. Devlin, Minal Patel, Antonin Berthon, Aijazuddin Syed, Mariko Yabe, Catherine Coombs, Nicole M. Caltabellotta, Mike Walsh, Ken Offit, Zsofia Stadler, Choonsik Lee, Paul Pharoah, Konrad H. Stopsack, Barbara Spitzer, Simon Mantha, James Fagin, Laura Boucai, Christopher J. Gibson, Benjamin Ebert, Andrew L. Young, Todd Druley, Koichi Takahashi, Nancy Gillis, Markus Ball, Eric Padron, David Hyman, Jose Baselga, Larry Norton, Stuart Gardos, Virginia Klimek, Howard Scher, Dean Bajorin, Eder Paraiso, Ryma Benayed, Maria Arcilla, Marc Ladanyi, David Solit, Michael Berger, Martin Tallman, Montserrat Garcia-Closas, Nilanjan Chatterjee, Luis Diaz, Ross Levine, Lindsay Morton, Ahmet Zehir, Elli Papaemmanuil. Memorial Sloan Kettering Cancer Center, New York, NY, University of North Carolina at Chapel Hill, Chapel Hill, NC, University of Cambridge, Cambridge, United Kingdom, Dana-Farber Cancer Institute, Boston, MA, Washington University, St Louis, MO, The University of Texas MD Anderson Cancer Center, Houston, TX, Moffitt Cancer Center, Tampa, FL, National Cancer Institute, Bethesda, MD

Recent studies among healthy individuals show evidence of somatic mutations in leukemia-associated genes, referred to as clonal hematopoiesis (CH). To determine the relationship between CH and oncologic therapy we collected sequential blood samples from 525 cancer patients (median sampling interval time = 23 months, range: 6-53 months) of whom 61% received cytotoxic therapy or external beam radiation therapy and 39% received either targeted/immunotherapy or were untreated. Samples were sequenced using deep targeted capture-based platforms. To determine whether CH mutational features were associated with tMN risk, we performed Cox proportional hazards regression on 9,549 cancer patients exposed to oncologic therapy of whom 75 cases developed tMN (median time to transformation=26 months). To further compare the genetic and clonal relationships between tMN and the proceeding CH, we analyzed 35 cases for which paired samples were available. We compared the growth rate of the variant allele fraction (VAF) of CH clones across treatment modalities and in untreated patients. A significant increase in the growth rate of CH mutations was seen in DDR genes among those receiving cytotoxic (p=0.03) or radiation therapy (p=0.02) during the follow-up period compared to patients who did not receive therapy. Similar growth rates among treated and untreated patients were seen for non-DDR CH genes such as DNMT3A. Increasing cumulative exposure to cytotoxic therapy (p=0.01) and external beam radiation therapy (2×10-8) resulted in higher growth rates for DDR CH mutations. Among 34 subjects with at least two CH mutations in which one mutation was in a DDR gene and one in a non-DDR gene, we studied competing clonal dynamics for multiple gene mutations within the same patient. The risk of tMN was positively associated with CH in a known myeloid neoplasm driver mutation (HR=6.9, p<10-6), and increased with the total number of mutations and clone size. The strongest associations were observed for mutations in TP53 and for CH with mutations in spliceosome genes (SRSF2, U2AF1 and SF3B1). Lower hemoglobin, lower platelet counts, lower neutrophil counts, higher red cell distribution width and higher mean corpuscular volume were all positively associated with increased tMN risk. Among 35 cases for which paired samples were available, in 19 patients (59%), we found evidence of at least one of these mutations at the time of pre-tMN sequencing and in 13 (41%), we identified two or more in the pre-tMN sample. In all cases the dominant clone at tMN transformation was defined by a mutation seen at CH Our serial sampling data provide clear evidence that oncologic therapy strongly selects for clones with mutations in the DDR genes and that these clones have limited competitive fitness, in the absence of cytotoxic or radiation therapy. We further validate the relevance of CH as a predictor and precursor of tMN in cancer patients. We show that CH mutations detected prior to tMN diagnosis were consistently part of the dominant clone at tMN diagnosis and demonstrate that oncologic therapy directly promotes clones with mutations in genes associated with chemo-resistant disease such as TP53.

  • therapy resulted also in clonal evolution and saw changes in splice variants and spliceosome
  • therapy promotes current DDR mutations
  • clonal hematopoeisis due to selective pressures
  • mutations, variants number all predictive of myeloid disease
  • deferring adjuvant therapy for breast cancer patients with patients in highest MDS risk group based on biomarkers, greatly reduced their risk for MDS

5704 – Pan-cancer genomic characterization of patient-matched primary, extracranial, and brain metastases

Presenter/AuthorsOlivia W. Lee, Akash Mitra, Won-Chul Lee, Kazutaka Fukumura, Hannah Beird, Miles Andrews, Grant Fischer, John N. Weinstein, Michael A. Davies, Jason Huse, P. Andrew Futreal. The University of Texas MD Anderson Cancer Center, TX, The University of Texas MD Anderson Cancer Center, TX, Olivia Newton-John Cancer Research Institute and School of Cancer Medicine, La Trobe University, AustraliaDisclosures O.W. Lee: None. A. Mitra: None. W. Lee: None. K. Fukumura: None. H. Beird: None. M. Andrews: ; Merck Sharp and Dohme. G. Fischer: None. J.N. Weinstein: None. M.A. Davies: ; Bristol-Myers Squibb. ; Novartis. ; Array BioPharma. ; Roche and Genentech. ; GlaxoSmithKline. ; Sanofi-Aventis. ; AstraZeneca. ; Myriad Genetics. ; Oncothyreon. J. Huse: None. P. Futreal: None.

Abstract: Brain metastases (BM) occur in 10-30% of patients with cancer. Approximately 200,000 new cases of brain metastases are diagnosed in the United States annually, with median survival after diagnosis ranging from 3 to 27 months. Recently, studies have identified significant genetic differences between BM and their corresponding primary tumors. It has been shown that BM harbor clinically actionable mutations that are distinct from those in the primary tumor samples. Additional genomic profiling of BM will provide deeper understanding of the pathogenesis of BM and suggest new therapeutic approaches.
We performed whole-exome sequencing of BM and matched tumors from 41 patients collected from renal cell carcinoma (RCC), breast cancer, lung cancer, and melanoma, which are known to be more likely to develop BM. We profiled total 126 fresh-frozen tumor samples and performed subsequent analyses of BM in comparison to paired primary tumor and extracranial metastases (ECM). We found that lung cancer shared the largest number of mutations between BM and matched tumors (83%), followed by melanoma (74%), RCC (51%), and Breast (26%), indicating that cancer type with high tumor mutational burden share more mutations with BM. Mutational signatures displayed limited differences, suggesting a lack of mutagenic processes specific to BM. However, point-mutation heterogeneity revealed that BM evolve separately into different subclones from their paired tumors regardless of cancer type, and some cancer driver genes were found in BM-specific subclones. These models and findings suggest that these driver genes may drive prometastatic subclones that lead to BM. 32 curated cancer gene mutations were detected and 71% of them were shared between BM and primary tumors or ECM. 29% of mutations were specific to BM, implying that BM often accumulate additional cancer gene mutations that are not present in primary tumors or ECM. Co-mutation analysis revealed a high frequency of TP53 nonsense mutation in BM, mostly in the DNA binding domain, suggesting TP53 nonsense mutation as a possible prerequisite for the development of BM. Copy number alteration analysis showed statistically significant differences between BM and their paired tumor samples in each cancer type (Wilcoxon test, p < 0.0385 for all). Both copy number gains and losses were consistently higher in BM for breast cancer (Wilcoxon test, p =1.307e-5) and lung cancer (Wilcoxon test, p =1.942e-5), implying greater genomic instability during the evolution of BM.
Our findings highlight that there are more unique mutations in BM, with significantly higher copy number alterations and tumor mutational burden. These genomic analyses could provide an opportunity for more reliable diagnostic decision-making, and these findings will be further tested with additional transcriptomic and epigenetic profiling for better characterization of BM-specific tumor microenvironments.

  • are there genomic signatures different in brain mets versus non metastatic or normal?
  • 32 genes from curated databases were different between brain mets and primary tumor
  • frequent nonsense mutations in TP53
  • divergent clonal evolution of drivers in BMets from primary
  • they were able to match BM with other mutational signatures like smokers and lung cancer signatures

5707 – A standard operating procedure for the interpretation of oncogenicity/pathogenicity of somatic mutations

Presenter/AuthorsPeter Horak, Malachi Griffith, Arpad Danos, Beth A. Pitel, Subha Madhavan, Xuelu Liu, Jennifer Lee, Gordana Raca, Shirley Li, Alex H. Wagner, Shashikant Kulkarni, Obi L. Griffith, Debyani Chakravarty, Dmitriy Sonkin. National Center for Tumor Diseases, Heidelberg, Germany, Washington University School of Medicine, St. Louis, MO, Mayo Clinic, Rochester, MN, Georgetown University Medical Center, Washington, DC, Dana-Farber Cancer Institute, Boston, MA, Frederick National Laboratory for Cancer Research, Rockville, MD, University of Southern California, Los Angeles, CA, Sunquest, Boston, MA, Baylor College of Medicine, Houston, TX, Memorial Sloan Kettering Cancer Center, New York, NY, National Cancer Institute, Rockville, MDDisclosures P. Horak: None. M. Griffith: None. A. Danos: None. B.A. Pitel: None. S. Madhavan: ; Perthera Inc. X. Liu: None. J. Lee: None. G. Raca: None. S. Li: ; Sunquest Information Systems, Inc. A.H. Wagner: None. S. Kulkarni: ; Baylor Genetics. O.L. Griffith: None. D. Chakravarty: None. D. Sonkin: None.AbstractSomatic variants in cancer-relevant genes are interpreted from multiple partially overlapping perspectives. When considered in discovery and translational research endeavors, it is important to determine if a particular variant observed in a gene of interest is oncogenic/pathogenic or not, as such knowledge provides the foundation on which targeted cancer treatment research is based. In contrast, clinical applications are dominated by diagnostic, prognostic, or therapeutic interpretations which in part also depends on underlying variant oncogenicity/pathogenicity. The Association for Molecular Pathology, the American Society of Clinical Oncology, and the College of American Pathologists (AMP/ASCO/CAP) have published structured somatic variant clinical interpretation guidelines which specifically address diagnostic, prognostic, and therapeutic implications. These guidelines have been well-received by the oncology community. Many variant knowledgebases, clinical laboratories/centers have adopted or are in the process of adopting these guidelines. The AMP/ASCO/CAP guidelines also describe different data types which are used to determine oncogenicity/pathogenicity of a variant, such as: population frequency, functional data, computational predictions, segregation, and somatic frequency. A second collaborative effort created the European Society for Medical Oncology (ESMO) Scale for Clinical Actionability of molecular Targets to provide a harmonized vocabulary that provides an evidence-based ranking system of molecular targets that supports their value as clinical targets. However, neither of these clinical guideline systems provide systematic and comprehensive procedures for aggregating population frequency, functional data, computational predictions, segregation, and somatic frequency to consistently interpret variant oncogenicity/pathogenicity, as has been published in the ACMG/AMP guidelines for interpretation of pathogenicity of germline variants. In order to address this unmet need for somatic variant oncogenicity/pathogenicity interpretation procedures, the Variant Interpretation for Cancer Consortium (VICC, a GA4GH driver project) Knowledge Curation and Interpretation Standards (KCIS) working group (WG) has developed a Standard Operating Procedure (SOP) with contributions from members of ClinGen Somatic Clinical Domain WG, and ClinGen Somatic/Germline variant curation WG using an approach similar to the ACMG/AMP germline pathogenicity guidelines to categorize evidence of oncogenicity/pathogenicity as very strong, strong, moderate or supporting. This SOP enables consistent and comprehensive assessment of oncogenicity/pathogenicity of somatic variants and latest version of an SOP can be found at https://cancervariants.org/wg/kcis/.

  • best to use this SOP for somatic mutations and not rearangements
  • variants based on oncogenicity as strong to weak
  • useful variant knowledge on pathogenicity curated from known databases
  • the recommendations would provide some guideline on curating unknown somatic variants versus known variants of hereditary diseases
  • they have not curated RB1 mutations or variants (or for other RBs like RB2? p130?)


Follow on Twitter at:








Read Full Post »

Metabolic Response Heterogeneity

Larry H Bernstein, MD, FCAP, Curator



The Prognostic Significance of Metabolic Response Heterogeneity in Metastatic Colorectal Cancer

PLoS One. 2015; 10(9): e0138341.

Published online 2015 Sep 30. doi:  10.1371/journal.pone.0138341

PMCID: PMC4589397

Alain Hendlisz,1,* Amelie Deleporte,1 Thierry Delaunoit,2 Raphaël Maréchal,3 Marc Peeters,4 Stéphane Holbrechts,6Marc Van den Eynde,7 Ghislain Houbiers,9 Bertrand Filleul,2 Jean-Luc Van Laethem,3 Sarah Ceyssens,5 Anna-Maria Barbuto,6 Renaud Lhommel,8 Gauthier Demolin,9 Camilo Garcia,10 Hazem El Mansy,1,2,3,4,5,6,7,8,9,10 Lieveke Ameye,11 Michel Moreau,11 Thomas Guiot,10 Marianne Paesmans,11 Martine Piccart,1 and Patrick Flamen10

Daniele Santini, Editor

Author information ► Article notes ► Copyright and License information ►

Go to:


Background Tumoral heterogeneity is a major determinant of resistance in solid tumors. FDG-PET/CT can identify early during chemotherapy non-responsive lesions within the whole body tumor load. This prospective multicentric proof-of-concept study explores intra-individual metabolic response (mR) heterogeneity as a treatment efficacy biomarker in chemorefractory metastatic colorectal cancer (mCRC).

Methods Standardized FDG-PET/CT was performed at baseline and after the first cycle of combined sorafenib (600mg/day for 21 days, then 800mg/day) and capecitabine (1700 mg/m²/day administered D1-14 every 21 days). MR assessment was categorized according to the proportion of metabolically non-responding (non-mR) lesions (stable FDG uptake with SUV-max decrease <15%) among all measurable lesions.

Results Ninety-two patients were included. The median overall survival(OS) and progression-free survival (PFS) were 8.2months (95%CI:6.8–10.5) and 4.2months (95%CI:3.4–4.8) respectively. In the 79 assessable patients, early PET-CT showed no metabolically refractory lesion in 47%, a heterogeneous mR with at least one non-mR lesion in 32%, and a consistent non-mR or early disease progression in 21%. On exploratory analysis, patients without any non-mR lesion showed a significantly longer PFS (HR 0.34; 95% CI: 0.21–0.56, P-value 0.02) compared to the other patients. The proportion of non-mR lesions within the tumor load did not impact PFS/OS.

Conclusion The presence of at least one metabolically refractory lesion is associated with a poorer outcome in advanced mCRC patients treated with combined sorafenib-capecitabine. Early detection of treatment-induced mR heterogeneity may represent an important predictive efficacy biomarker in mCRC.

Trial Registration ClinicalTrials.gov NCT01290926



Tumoral heterogeneity is a major determinant of resistance in solid tumors. FDG-PET/CT can identify early during chemotherapy non-responsive lesions within the whole body tumor load. This prospective multicentric proof-of-concept study explores intra-individual metabolic response (mR) heterogeneity as a treatment efficacy biomarker in chemorefractory metastatic colorectal cancer (mCRC).


Standardized FDG-PET/CT was performed at baseline and after the first cycle of combined sorafenib (600mg/day for 21 days, then 800mg/day) and capecitabine (1700 mg/m²/day administered D1-14 every 21 days). MR assessment was categorized according to the proportion of metabolically non-responding (non-mR) lesions (stable FDG uptake with SUVmax decrease <15%) among all measurable lesions.


Ninety-two patients were included. The median overall survival (OS) and progression-free survival (PFS) were 8.2 months (95% CI: 6.8–10.5) and 4.2 months (95% CI: 3.4–4.8) respectively. In the 79 assessable patients, early PET-CT showed no metabolically refractory lesion in 47%, a heterogeneous mR with at least one non-mR lesion in 32%, and a consistent non-mR or early disease progression in 21%. On exploratory analysis, patients without any non-mR lesion showed a significantly longer PFS (HR 0.34; 95% CI: 0.21–0.56, P-value <0.001) and OS (HR 0.58; 95% CI: 0.36–0.92, P-value 0.02) compared to the other patients. The proportion of non-mR lesions within the tumor load did not impact PFS/OS.


The presence of at least one metabolically refractory lesion is associated with a poorer outcome in advanced mCRC patients treated with combined sorafenib-capecitabine. Early detection of treatment-induced mR heterogeneity may represent an important predictive efficacy biomarker in mCRC.

Trial Registration

ClinicalTrials.gov NCT01290926


The development of new therapeutics for solid tumors is currently strained by increasing regulatory demands to better define subpopulations bearing resistant diseases in order to spare patients from useless toxicities and the society from unaffordable costs in case of ineffective treatments.

Tumor heterogeneity through the existence of resistant subclones (genetic drift) or local environmental factors is nowadays accepted as a major determinant of treatment resistance. However, sensitive biomarkers of tumoral heterogeneity are lacking.[13] Current response assessment methods using morphology (RECIST using MRI/CT) or metabolism (PERCIST using FGD-PET/CT) do not allow the description of tumor heterogeneity because dichotomization of response (versus non-response) requires summing of measurements or the selection of the one single most representative lesion.[4] Moreover most of the new biological therapies render response evaluation even more challenging by the infrequency of tumor shrinkage.[58]

Imaging tumour metabolism using 18F-Fluorodeoxyglucose positron emission tomography coupled with computed tomography (FDG-PET/CT) allows rapid identification of treatment-refractory lesions with a high negative predictive value (NPV).[914] FDG-PET is currently central in the international recommendations for response assessment for Hodgkin’s disease and aggressive non-Hodgkin’s lymphoma, in which medical conditions it is used commonly as a basis for therapeutic decisions. [1417] In contrast, solid tumors are frequently more refractory to treatment and reveal smaller and slower changes in FDG uptake under therapy leading to the existence of different criteria for metabolic response assessment at the lesion as well as at the patient level.[18,19] This ongoing discussion explain why metabolic imaging has still not acquired a biomarker status in solid tumors.

Metabolic imaging provides a whole-body quantitative assessment of treatment-induced changes in tumoral glycolysis early after treatment initiation, before any morphological changes are observed. It has therefore the potential to detect tumoral heterogeneity by revealing how distinct tumor sites behave in response to treatment.

Several trials suggest meaningful clinical activity of combined sorafenib-capecitabine in metastatic breast and colorectal cancer. However the significant toxicity of the combination renders its use practically incompatible with a palliative setting, further underscoring the need to identify a sensitive biomarker for patient selection.[20,21] Preliminary reports in lung and renal cancer suggest that FDG-PET-based metabolic response assessment could be used as a predictive biomarker of sorafenib.[22,23]

The trial is a proof-of-concept study designed to explore intra-individual mR heterogeneity as a prognostic biomarker for this combination of a biological and a cytotoxic agent in mCRC.



Material and Methods

Belgian competent authorities and ethical committees of the 6 participating centres approved the study (EudraCT 2010-023695-91, clinicaltrials.gov NCT01290926), designed as a prospective multicentric single-arm phase II, with one-stage accrual.

Patients with histologically proven unresectable metastatic CRC failing all standard treatments but not necessarily bevacizumab were eligible. Exclusion criteria were contraindications for capecitabine and sorafenib, ECOG performance status (PS) > 1, age < 18 years, and cerebral metastasis. Normal organ and bone marrow function, a life expectancy >12 weeks, and a signed informed consent were required.

Both drugs were given orally on an outpatient basis: sorafenib 200mg in the morning and 400 mg in the evening every day for the first cycle, then 400 mg twice a day every day; capecitabine 850 mg/m2 twice a day on days 1 to 14, every 21 days. One cycle was defined as a 21-day period. Adverse events were reported according to the National Cancer Institute Criteria, version 3.0 (http://ctep.cancer.gov/protocolDevelopment/electronic_applications/docs/ctcaev3.pdf). Study medications were to be stopped at disease progression or when unacceptable toxicity occurred. RECIST 1.1-radiological response was assessed locally every two cycles (6weeks). Patients were followed until objective disease progression and every 3 months thereafter for survival assessment.

FDG-PET/CT Imaging

For the FDG-PET/CT, patients were referred to one of the 5 participating PET/CT centres, previously approved for participation based on FDG-PET phantom imaging study for quality’s central assessment [24]. An independent academic molecular imaging core laboratory (OriLab) centralized all FDG-PET/CT images through anonymized CD-Rom transfers, checked image’s quality, DICOM headers, compliance to the Standard Procedures Imaging Manual and imaging case report forms.

Baseline FDG-PET/CT was performed within 7 days preceding chemotherapy initiation and repeated under the same technical and patient conditions on day (D)21 (range D19-D23), with D1 as the first day of chemotherapy administration. Prior to FDG injection, fasting ≥ 6 hours and glycemia levels <120 mg/dL for non-diabetic patients, and <180 mg/dL for diabetic patients were required. Short-acting insulin use on the day of FDG-PET/CT was not allowed.

The PET/CT was initiated 60 to 90 minutes after intravenous injection of 3.7 to 7.4 MBq/kg FDG, optimized for body weight. Similar FDG activity (+/-15%) and time window (+/- 15 min) were used for the second PET/CT.

Whole body scanning with a low dose CT (without intravenous or oral contrast, from proximal femur to skull) was performed, immediately followed by the PET acquisition. Imaging acquisition and reconstruction remained stable over the whole study period. The second FDG-PET/CT was strictly blinded to the investigators, and was not added to the patient’s (electronic) medical records.

The standard uptake value (SUV) of FDG used was the lean body mass-based maximal SUV value within the lesion (SUVmax, g/ml).

All FDG-PET/CT images were analysed in batches using the same software (PETVcar version 4.6, General Electric, USA) and display techniques. Two senior nuclear medicine physicians (PF, CG) performed independent mR analyses using a predefined 3-step methodology.[13] First, on the baseline PET/CT, target lesions were identified according to the following criteria: transaxial diameter (measured on the CT of the PET/CT) > 15 mm, intense FDG uptake (> 2 x normal liver parenchym uptake) and with an unequivocally neoplastic basis. Each target lesion was then classified as non-responding (decrease of SUVmax on second PET-CT<15%) or responding. Second, the patients were classified according to the lesional distribution of mR; class I: absence of any metabolically non-responding lesion, class II: a minor part of whole body tumour load shows a non-response, class III: major part of whole body target tumour load does not respond, and, class IV: all target lesions are non-responding, or presence of a progressive lesion (progression defined as >25% increase of SUVMax, or appearance of a new lesion). (Fig 1) Finally, different methods of patient response dichotomization (metabolic responders versus non-responders) were explored.

Fig 1

Classes of metabolic responses. Class1: no metabolic unresponsive lesion; Class2: minority of unresponsive lesions among whole body target tumour load; Class3: majority of whole body target tumour load does not respond; Class4: all target lesions are non-responding,or, presence of progressive lesions [progression defined as > 25% increase of FDG up take on second PET or appearance of a new lesion]  http://dx.doi.org:/10.1371/journal.pone.0138341.g001


Classes of metabolic responses.

Statistical considerations

A first co-primary objective defined the minimal clinical activity necessary to explore the negative predictive value of metabolic response imaging on OS as a survival rate at 6 months > 30% according to the existing literature on chemorefractory CRC. To reject the null hypothesis that the 6 month-OS rate would be <30% using a binomial distribution, a 1-sided test with α = 0.025 and a power of 90% in case of a true 6 months-OS ≥ 50% was used, requiring a sample size of 66 eligible patients followed for at least 6 months. An intent-to-treat (ITT) approach was used.

The second co-primary objective was the prognostic value of mR classification. Based on a previous study,[13] and anticipating a 95% eligibility rate, a 50% early PET/CT non-responders rate, and a hazard ratio (HR) around 0.385 for comparison between the survival distributions, 54 events were needed for a 90% power and a two-sided logrank test at the 2.5% level.

Because the mR rate monitored during the study was higher than expected, the number of events to be observed was increased to 62. This decision was taken without changing the HR to be detected and without estimating this HR during study conduct.

Secondary objectives were to describe PFS, objective response rate and toxicity and to determine the predictive value of early MR on PFS.

For the first co-primary objective, the 6 month-OS, median (m)OS and mPFS were calculated from the patient’s inclusion. For the second co-primary objective, the predictive value assessment of mR on OS and PFS was done from the time of the second FDG-PET/CT on patients having undergone the second FDG-PET/CT in order to control for guarantee-time bias.[25] PFS was calculated up to the time of disease progression or death, whichever occurred first. Kaplan-Meier estimates were used to characterise PFS and OS, and the log-rank test to investigate comparisons between survival curves. Cox’s proportional hazards model was used to calculate HR and their 95% CI

The multivariate analysis was performed using Cox’s proportional hazard model. Variables with a univariate P-value < 0.20 were considered as possible predictors in the multivariate model. We performed stepwise forward selection of variables, i.e. forward selection but at each step variables already in the model could be dropped if their associated p-value became >0.05. To verify the final model, also backward selection of variables was performed on all variables with univariate p-value<0.20, resulting in the same set of variables.[26]

All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and GraphPad Prism 6 software.

Patients found with an early metabolic progressive disease (class IV) were not excluded from the statistical analyses as the objectives of the paper were to show the predictive value of early metabolic response on OS and PFS, which implies the necessity of an intent-to-treat analysis. The event “progression” in the definition of PFS is moreover a radiological progression. Patients belonging to class IV do not meet this definition of radiological progression, which remains an event to be predicted.


Between February and October 2011, 97 consecutive patients were enrolled in 6 clinical centres. The CONSORT diagram details the reasons for considering 5 patients as ineligible, excluding them from all analysis (Fig 2). The eligible patients (N = 92), median age 63 (range 28–83), male/female ratio of 54/46, PS 0 (55%) or 1(45%) received a median of 5 (range 0–44+) cycles of sorafenib-capecitabine after an history of a median of 3 (range 1–6) prior therapeutic lines including bevacizumab in 55% of patients. Codons 12–13 KRAS mutations were present in 52%.

Fig 2

Consort Diagram.  http://dx.doi.org:/10.1371/journal.pone.0138341.g002

Toxicity (Table 1)

Table 1

Most important (>10%) side effects in the 88 patients who received treatment according to Common Toxicity Criteria CTC3.0.  http://dx.doi.org:/10.1371/journal.pone.0138341.t001

Patients presented a median of 7 (Q1 = 4, Q3 = 9) different adverse reactions during therapy. All but one patient experienced at least one toxicity of any grade, of whom 61.4% with at least one grade III-IV. Grade III-IV side effects were mainly fatigue (21.6%), hand-foot skin reactions (HFSR) (15.9%), and diarrhoea (12.5%). No toxic death was observed. Toxicity led to dose modifications in 63.6% and therapy discontinuation in 5.7% of cases.

Survival data and radiological response

The mOS and mPFS were 8.2 months (95% CI: 6.8–10.5) and 4.2 months (95% CI: 3.4–4.8) respectively. The OS rate at 6 months was 71% (65/92) (95% CI: 61%-79%), significantly higher than the 30% minimal efficiency level predefined in the statistical plan (p-value <0.001), meeting the clinical co-primary endpoint.

According to RECIST, partial response was observed in 7/92 patients (7.6%, 95%CI 2.2–13.0). In the 79 assessable patients, disease control at first evaluation (partial responses and stable diseases according to RECIST) was noted in 32/37 (80%) of the patients with consistent mR versus 24/42 (57%) in other patients (p-value 0.006) (Table 2).

Table 2

RECIST1.1 versus Metabolic Response classes in patients for whom both mR and RECIST assessment of response are available.  http://dx.doi.org:/10.1371/journal.pone.0138341.t002

Metabolic response analysis

MR data were available for 79 patients: 37 (46.8%) were classified as class I; 14 (17.7%) as class II; 11 (13.9%) as class III; and 17 (21.5%) as class IV. Within Class IV, 8 patients (10%) showed early metabolic disease progression.

Patients without any metabolically non-responding lesions (Class I) performed better than patients with heterogeneous responses (Class II and III) or with a consistent non-response or progressive disease (Class IV). The difference between the four classes is statistically significant for mPFS (p-value <0.001) but not for mOS (p-value = 0.13). (Fig 3A and 3B)

Fig 3

PFS* (A) and OS* (B) distribution according to the 4 classes of metabolic response.  Class1: no metabolic unresponsive lesion; Class2: minority of unresponsive lesions among whole body target tumour load; Class3: majority of whole body target tumour load does not respond; Class 4: all target lesions are nonresponding, or, presence of progressive lesions [progression defined as >25% increase of FDG uptake on second PET, or appearance of a new lesion].*from date of the second FDG PET-CT.

Two classifications were considered for reporting response in a dichotomized way according to mR heterogeneity among lesions: classes (I and II) versus classes (III and IV),[13] and classes (I) versus classes (II+III+IV). The first compares outcome according to the dominance of non-mR lesions within the tumor load, the second according to the consistence of mR (Table 3Fig 4). “Using the “dominance” classification to define early metabolic non response, the second co-primary objective, which was to identify a prognostic value on survival for early metabolic assessment, was not met while it was successful to discriminate patients according to their outcome using the exploratory “consistence” classification.“Five of the 42 patients (12%) with at least one non-responding lesion remained free of disease progression at 6 months, versus 15 of the 37 class I patients (41%) (p-value 0.005).



Table 3 Correlation of mPFS and mOS with Dominance and Consistency of metabolic response.  http://dx.doi.org:/10.1371/journal.pone.0138341.t003


Fig 4   PFS and OS distribution according to the dichotomized mR classifications.  http://dx.doi.org:/10.1371/journal.pone.0138341.g004

Multivariate analysis after stepwise variable selection of age, PS, number of previous chemotherapy lines, bevacizumab pretreatment, sex, Body Mass Index (BMI), HFSR occurrence and mR retained the absence of metabolically resistant lesion (class I) as the only variable significantly correlated with both mOS and mPFS (Table 4).

Table 4  Univariate and multivariate analysis for OS and PFS.   http://dx.doi.org:/10.1371/journal.pone.0138341.t004


Tumoral heterogeneity, described as the coexistence of genomically different subclones within a patient tumor load or to local environmental aspects, is recognized as a major determinant of resistance to treatment in solid tumors.[13] However, interlesional tumor heterogeneity in metastatic setting is not covered by current response assessment methods because of the analysis’ methodology performing averaging of responses among lesions. This prospective multicentric proof-of-concept study explored interlesional mR heterogeneity as a biomarker of treatment resistance in advanced solid tumors.

As previously reported in several solid tumors, FDG-PET/CT response assessment after one therapy cycle allows a rapid identification of non-responding lesions/patients, fulfilling the necessary conditions to become potentially a good predictive biomarker, which is crucial to avoid useless toxicity.[4,912,22,27] Moreover, significant progresses and implementation of standardized methodology for FDG-PET/CT imaging, including homogenization of imaging procedures and patient’s preparation, quality control and independent central analysis, now allows its use in multicentric trials.[24,27,28]

Studying tumoral heterogeneity requires assessing the response of the whole baseline metastatic tumor load without restriction in number nor site. However, existing morphological (WHO, RECIST) and metabolic (EORTC, PERCIST) response assessment methods do not take into account this response heterogeneity because they only consider a limited number of operator-selected target lesions and/or perform summing or averaging of response variables.[4,19,29,30] Moreover, being classically performed late during treatment, these assessment criteria measure response, while from a clinical point of view, it is the presence of non-response that triggers the need for treatment adaptation. For this, based on prior research, in order to optimize the negative predictive value (NPV) of mR assessment, a 15% cut-off value of SUVmax decrease instead of the standard 25–30% response cut-off value was chosen.[18,31] Such low cut-off value maximally avoids unjustified denial to a potentially active treatment regimen.

With regard to the characterization of response heterogeneity among lesions, this study adopted a multistep descriptive procedure. First, a lesion-by-lesion response analysis of all measurable lesions on baseline FDG-PET/CT without restriction of their number was performed applying the 15% cut-off for non-response. Then, a patient-based 4-class classification was applied, describing the presence and proportion of metabolically non-responding lesions among the whole-body tumor load.[13]

Using such methodology, 22% of the patients showed overall treatment resistance of whom 10% showed early metabolic disease progression at 3 weeks. This observation indicates the importance of performing the baseline FDG-PET/CT as close as possible before the start of the tested drug administration, because rapid disease progression during this timeframe could lead to false negative mR assessment.

On the other hand, after one treatment cycle, 32% of the patients showed heterogeneous metabolic responses combining resistant with potentially responding lesions (Class II and III). Of these, 18% showed non-mR in the minor, while 14% showed a non-mR in the major part of the tumor load. The proportion of heterogeneous response observed in this study is considerable, confirming earlier observation in an independent mCRC patient group treated with chemotherapy, where heterogeneity of mR was described in 67% of patients.[13] Other comparisons are impossible because information about heterogeneity is lacking in most available literature, which apply dichotomization to response assessment.[3234]

Indeed, for clinical decision-making, the response assessment is generally reported dichotomously, because clinicians have to decide whether to continue or adapt the initiated treatment. Such information-reducing response reporting may only be adequate in case of homogeneous mR, but blurs useful information in case of response heterogeneity.

Outcome analysis in this study indicated that mPFS and mOS are comparable in patients bearing one or more metabolically resistant lesion. Only patients without any resistant lesion (class I) seemed to have a better outcome (mPFS and mOS) compared to all others. Therefore it seems that the presence but not the number/proportion of non-responding lesions is the most important prognostic determinant. Moreover, its value is reinforced by a multivariate analysis showing absence of any metabolically treatment resistant lesion as an independent prognostic factor for both PFS and OS.

A valid assessment of a predictive biomarker requires a significant level of activity of the regimen under study. This was achieved, as 71% of the included patients were still alive at 6 months, which was significantly higher than the minimal activity predefined in the study design. ITT analysis of the 92 eligible patients showed a mPFS of 4.2 months and a mOS of 8.2 months respectively, suggesting an overall beneficial effect for this drug combination compared to recent historical data with 2 months mPFS and 4–6 months mOS in the same clinical setting.[6,31,3537]

Moreover, this study confirms the need for an effective predictive response biomarker for a sorafenib-containing regimen, because of the high toxicity profile together with the poor sensitivity of morphology-based imaging (CT/MRI) for detecting responses (only 8% of partial response according to RECIST) during treatment.[7,8,38]

A major application of standardized metabolic imaging is expected in early drug development (phase I-II) for two reasons: (i) as FDG-PET response analysis seems to be correlated with prognosis, it provides a rapid appraisal of the new drug activity even in small patient populations, and (ii) image-guided biopsies of resistant lesions could identify the molecular basis of treatment resistance by demonstrating genomic or epigenomic heterogeneity.

In this study for instance, half (47%) of the patients didn’t demonstrate any resistant lesion, indicating a remarkable activity level for such a heavily pre-treated patients population, unsuspected by classical morphological imaging.

Furthermore, in the metastatic setting, FDG-PET/CT may provide a tool for the identification of patients with one or very few metastatic sites resisting to treatment for whom the continuation of unchanged therapy carries a grim prognosis. This raises the potential of adding locoregional ablative treatments guided by the imaging of metabolic response, in order to achieve homogeneity of disease control and restore prognosis. If the current observation is confirmed by an ongoing multicentric trial, (clinicaltrials.gov NCT01929616), randomized prospective trials using early FDG-PET/CT response assessment as an interventional tool for targeting locoregional therapy (eg. surgery, radioembolization, radiofrequency ablation) will be justified.

Finally, in the absence of randomized data based on PET response, it remains to be proven whether the presence of metabolically non-responding lesions is a biomarker identifying more heterogeneous diseases with intrinsically a worse prognosis, or a genuine therapeutic predictive tool for a given treatment.



Metabolic response assessment allows the early identification of treatment-resistant tumor sites. The presence of metabolically refractory lesions seems to negatively impact overall treatment outcome whatever their number, adding to the mounting evidence that tumour heterogeneity is a key element in cancer management.

Early assessment of mR heterogeneity is a potentially powerful predictive biomarker enabling the personalization of anticancer treatments by increasing their cost-effectiveness and sparing useless toxicities.

Go to:

Supporting Information

S1 Protocol

Study protocol.


Click here for additional data file.(1.1M, pdf)

S1 TREND Checklist

TREND Checklist.


Click here for additional data file.(1.3M, pdf)


  1. McDermott U, Downing JR, Stratton MR (2011) Genomics and the continuum of cancer care. N Engl J Med 364: 340–350. doi: 10.1056/NEJMra0907178[PubMed]
  2. Greaves M, Maley CC (2012) Clonal evolution in cancer. Nature 481: 306–313. doi:10.1038/nature10762[PMC free article] [PubMed]
  3. Aparicio S, Caldas C (2013) The implications of clonal genome evolution for cancer medicine. N Engl J Med 368: 842–851. doi: 10.1056/NEJMra1204892[PubMed]
  4. Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. (2009) NSew response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 45: 228–247.[PubMed]
  5. Grothey A, Hedrick EE, Mass RD, Sarkar S, Suzuki S, Ramanathan RK, et al. (2008) Response-independent survival benefit in metastatic colorectal cancer: a comparative analysis of N9741 and AVF2107.J Clin Oncol 26: 183–189. doi: 10.1200/JCO.2007.13.8099[PubMed]
  6. Grothey A, Van Cutsem E, Sobrero A, Siena S, Falcone A, Ychou M, et al. (2012) Regorafenib monotherapy for previously treated metastatic colorectal cancer (CORRECT): an international, multicentre, randomised, placebo-controlled, phase 3 trial. Lancet 381: 303–312. doi: 10.1016/S0140-6736(12)61900-X[PubMed]
  7. Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. (2008) Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 359: 378–390. doi: 10.1056/NEJMoa0708857[PubMed]
  8. Awada A, Gil T, Whenham N, Van Hamme J, Besse-Hammer T, Brendel E, et al. (2011) Safety and pharmacokinetics of sorafenib combined with capecitabine in patients with advanced solid tumors: results of a phase 1 trial. J Clin Pharmacol 51: 1674–1684. doi: 10.1177/0091270010386226[PubMed]
  9. Ott K, Weber WA, Lordick F, Becker K, Busch R, Herrmann K, et al. (2006) Metabolic imaging predicts response, survival, and recurrence in adenocarcinomas of the esophagogastric junction. J Clin Oncol 24: 4692–4698. [PubMed]
  10. Hoekstra CJ, Stroobants SG, Smit EF, Vansteenkiste J, van Tinteren H, Postmus PE, et al. (2005)Prognostic relevance of response evaluation using [18F]-2-fluoro-2-deoxy-D-glucose positron emission tomography in patients with locally advanced non-small-cell lung cancer. J Clin Oncol 23: 8362–8370. [PubMed]
  11. Rousseau C, Devillers A, Sagan C, Ferrer L, Bridji B, Campion L, et al. (2006) Monitoring of early response to neoadjuvant chemotherapy in stage II and III breast cancer by [18F]fluorodeoxyglucose positron emission tomography. J Clin Oncol 24: 5366–5372. [PubMed]
  12. de Geus-Oei LF, van Laarhoven HW, Visser EP, Hermsen R, van Hoorn BA, Kamm YJ, et al. (2008)Chemotherapy response evaluation with FDG-PET in patients with colorectal cancer. Ann Oncol 19: 348–352. [PubMed]
  13. Hendlisz A, Golfinopoulos V, Garcia C, Covas A, Emonts P, Ameye L, et al. (2012) Serial FDG-PET/CT for early outcome prediction in patients with metastatic colorectal cancer undergoing chemotherapy.Ann Oncol 23: 1687–1693. doi: 10.1093/annonc/mdr554[PubMed]
  14. Barrington SF, Mikhaeel NG, Kostakoglu L, Meignan M, Hutchings M, Mueller SP, et al. (2014) Role of Imaging in the Staging and Response Assessment of Lymphoma: Consensus of the International Conference on Malignant Lymphomas Imaging Working Group. J Clin Oncol. [PubMed]
  15. Cheson BD, Pfistner B, Juweid ME, Gascoyne RD, Specht L, Horning SJ, et al. (2007) Revised response criteria for malignant lymphoma. J Clin Oncol 25: 579–586. [PubMed]


Read Full Post »

Notes On Tumor Heterogeneity: Targets and Mechanisms, from the 2015 AACR Meeting in Philadelphia PA

Reporter: Stephen J. Williams, Ph.D.

The following contain notes from the Sunday April 19, 2015 AACR Meeting (Pennsylvania Convention Center, Philadelphia PA) 1 PM Major Symposium Session on Tumor Heterogeneity: Targets and Mechanism chaired by Dr. Charles Swanton.

Speakers included: Mark J. Smyth, Charles Swanton, René H. Medema, and Catherine J. Wu

Tumor heterogeneity is a common feature of many malignancies, especially the solid tumors and can drive the evolution and adaptation of the growing tumor, complicating therapy and resulting in therapeutic failure, including resistance. This session at AACR described the mechanisms, both genetic and epigenetic, which precipitate intratumor heterogeneity and how mutational processes and chromosomal instability may impact the tumor progression and the origin of driver events during tumor evolution. Finally the session examined possible therapeutic strategies to take advantage of, and overcome, tumor evolution. The session was chaired by Dr. Charles Swanton. For a more complete description of his work, tumor heterogeneity, and an interview on this site please click on the link below:

Issues in Personalized Medicine in Cancer: Intratumor Heterogeneity and Branched Evolution Revealed by Multiregion Sequencing


Issues in Personalized Medicine: Discussions of Intratumor Heterogeneity from the Oncology Pharma forum on LinkedIn


Notes from Charles Swanton, Cancer Research UK; Identifying Drivers of Cancer Diversity

Dr. Swanton’s lecture focused on data from two recent papers from his lab by Franseco Favero and Nicholas McGranahan:

  1. Glioblastoma adaptation Traced Through Decline of an IDH1 clonal driver and macro-evolution of a double-minute chromosome (Annals of Oncology, 2015)[1]

This paper described the longitudinal Whole Genome Sequencing (WGS) study of a 35 year old female whose primary glioblastoma (GBM) was followed through temozolomide treatment and ultimately recurrence.

  • In 2008 patient was diagnosed with primary GBM (three biopsies of unrelated sites were Grade II and Grade IV; temozolomide therapy for three years then relapse in 2011
  • WGS of 2 areas of primary tumor showed extensive mutational and copy number heterogeneity; was able to identify clonal TP53 mutations and clonal IDH1 mutation in primary tumor with different patterns of clonality based on grade
  • Amplifications on chromosome 4 and 12 (PDGFRA, KIT, CDK4)
  • After three years of temozolomide multiple translocations found in chromosome 4 and 12 (6 translocations)
  • Clonal IDH1 R132H mutation in primary tumor only at very low frequency in recurrent tumor
  • The WGS on recurrent tumor (sequencing took ONLY 9 days from tumor resection to sequence results) showed mutation cluster in KIT/PDGFRA.PI3K.mTOR axis so patient treated with imatinib
  • However despite rapid sequencing and a personalized approach based on WGS results, tumor progressed and patient died shortly: tumor evolution is HUGE hurdle for personalized medicine

As Dr. Swanton stated:

“we are underestimating the frequency of polyclonal evolution”

  1. Clonal status of actionable driver events and the timing of mutational processes in cancer evolution (Science Translational Medicine, 2015)[2]
  • analyzed nine cancer types to determine the subclonal frequencies of driver events, to time mutational processes during cancer evolution, and to identify drivers of subclonal expansions.
  • identified later subclonal “actionable” mutations, including BRAF (V600E), IDH1 (R132H), PIK3CA (E545K), EGFR (L858R), and KRAS (G12D), which may compromise the efficacy of targeted therapy approaches.
  • > 20% of IDH1 mutations in glioblastomas, and 15% of mutations in genes in the PI3K (phosphatidylinositol 3-kinase)–AKT–mTOR (mammalian target of rapamycin) signaling axis across all tumor types were subclonal
  • Mutations in the RAS–MEK (mitogen-activated protein kinase kinase) signaling axis were less likely to be subclonal than mutations in genes associated with PI3K-AKT-mTOR signaling

Branched chain can converge on single resistance mechanism; clonal resistance (for example to PI3K inhibitors can get multiple PTEN mutations in various metastases

Targeting Tumor Heterogeneity

  • Identify high risk occupants (have to know case history)
  • Mutational landscape interferes with anti-PD1 therapies
  • Low frequency mutations affect outcome

Notes from Dr. Catherine J. Wu, Dana-Farber Cancer Institute: The evolutionary landscape of CLL: Therapeutic implications

  • Clonal evolution a key feature of cancer progression and relapse
  • Hypothesis: evolutionary dynamics (heterogeneity) in chronic lymphocytic leukemia (CLL) contributes to variations in response and disease “tempo”
  • Used whole exome sequencing and copy number data of 149 CLL cases to discover early and late cancer drivers: clonal patterns (Landau et. al, Cell 2013); some drivers correspond to poor clinical outcome
  • Methylation studies suggest that there is epigenetic heterogeneity which may drive CLL clonal evolution
  • Developing methodology to integrate WES to determine mutations with immunogenic potential for development of personalized immunotherapy for CLL and other malignancies


  1. Favero F, McGranahan N, Salm M, Birkbak NJ, Sanborn JZ, Benz SC, Becq J, Peden JF, Kingsbury Z, Grocok RJ et al: Glioblastoma adaptation traced through decline of an IDH1 clonal driver and macro-evolution of a double-minute chromosome. Annals of oncology : official journal of the European Society for Medical Oncology / ESMO 2015, 26(5):880-887.
  2. McGranahan N, Favero F, de Bruin EC, Birkbak NJ, Szallasi Z, Swanton C: Clonal status of actionable driver events and the timing of mutational processes in cancer evolution. Science translational medicine 2015, 7(283):283ra254.


Other related articles on Tumor Heterogeneity were published in this Open Access Online Scientific Journal, include the following:


Issues in Personalized Medicine: Discussions of Intratumor Heterogeneity from the Oncology Pharma forum on LinkedIn

Issues in Personalized Medicine in Cancer: Intratumor Heterogeneity and Branched Evolution Revealed by Multiregion Sequencing

CANCER COMPLEXITY: Heterogeneity in Tumor Progression and Drug Response – 2015 Annual Symposium @Koch Institute for Integrative Cancer Research at MIT – W34, 6/12/2015 9:00 AM EDT – 4:30 PM EDT

My Cancer Genome from Vanderbilt University: Matching Tumor Mutations to Therapies & Clinical Trials

Tumor Imaging and Targeting: Predicting Tumor Response to Treatment: Where we stand?

Mitochondrial Isocitrate Dehydrogenase and Variants

War on Cancer Needs to Refocus to Stay Ahead of Disease Says Cancer Expert

Read Full Post »