- Email: [email protected]
Towards a screening test for cancer by circulating DNA analysis
abstracts
1425P
Towards a screening test for cancer by circulating DNA analysis
R. Tanos1, A. Otandault1, C. Mollevi1, A. Bauer1, G. Tousch1, L. Picque Lasorsa1, S. El Messaoudi1, J. Colinge1, P-E. Colombo1, W. Jacot1, T. Mazard1, J.M. Sayague´s2, B. Gillet3, D. Pezet3, M. Ychou1, A.R. Thierry1 1 IRCM (Institute of Cancer Research of Montpellier), ICM Regional Cancer Institute of Montpellier, Montpellier, France, 2Department of Medicine and Cancer Research Center, University Hospital, Salamanca, Spain, 3Oncologie Digestive et Dermatologie, CHU Clermont-Ferrand, Clermont-Ferrand, France Background: Circulating DNA (cfDNA) has emerged as a potential biomarker in cancer, and is the subject of extensive studies in translational and clinical research. Our group has been interested in its implication and clinical significance in the field of oncology for many years, and is now focused on evaluating its potential for early cancer detection. Methods: We recently developed a screening test (MNR: Multi Normalized Ratio), based on various cfDNA parameters determined by a specific q-PCR based method, targeting both nuclear and mitochondrial sequences. Results: When applied to the supernatant of cell culture, the MNR had a discriminative potential of 100% between normal and cancer cell lines. An extensive evaluation of this test was carried out in plasma samples of 289 healthy subjects and 987 cancer patients (CRC, breast, liver, pancreatic, ovarian) of all stages. Preliminary results revealed a high potential with an AUC of 0.81 (0.78-0.84, 95% CI), a 70% sensitivity (Se) and 77% specificity (Sp). In breast cancer (N ¼ 169), an AUC of 0.82 (0.78-0.86, 95% CI) with 72% Se and 80% Sp were observed. In all stages CRC patients (N ¼ 795), the results showed an AUC of 0.80 (0.78-0.84, 95%, CI), 75% Se and 70% Sp; for CRC stages 0/I/II (N ¼ 426), an AUC of 0.79 (0.75-0.82, 95% CI), 70% Se and 72% Sp; and for CRC stage IV (N ¼ 186), a 0.86 AUC (0.82-0.89, 95% CI) with 75% Se and 80% Sp. When combining the MNR to a total cfDNA concentration threshold value (AUC ¼ 0.81 (0.790.83, 95% CI), 72% Se and 76% Sp for all stage cancers (N ¼ 987)), in a test cohort of 173 stages 0/I/II CRC patients and 132 healthy individuals, we increased the sensitivity and specificity to 74% and 95% respectively. Furthermore we recently discovered that cfDNA fragmentation, as determined by Whole Genome Sequencing using either double or single strand library, is also a parameter enabling discrimination between healthy and cancer individuals. Conclusions: The implementation of a multi-parametric test combining total cfDNA quantification, MNR and fragmentation biomarkers, with help of a decision tree in machine learning, is currently on-going. Our data suggest that our strategy in targeting cfDNA structural features might be powerful for early cancer detection, and appears as an alternative or a synergistic combination to the detection of mutations. Legal entity responsible for the study: Alain R. Thierry - INSERM (Institut national de la sante´ et de la recherche me´dicale). Funding: MSDAvenir - Mitest / Alain R. Thierry is supported by INSERM (Institut national de la sante´ et de la recherche me´dicale). Disclosure: All authors have declared no conflicts of interest.
v580 | New Diagnostic Tools
1426P
Evaluation of a successful launch of the MammaPrint and BluePrint NGS kit
L. Delahaye1, A.T. Witteveen1, M. Snel1, T. Cavness2, B. Chan3, L. Mittempergher1, A.M. Glas1 1 Product Development, Agendia Inc., Amsterdam, Netherlands, 2Operations, Agendia Inc., Irvine, CA, USA, 3Product Support, Agendia Inc., Irvine, CA, USA Background: Centralized MammaPrint (MP) and BluePrint (BP) microarray-based genomic tests on FFPE RNA were succesfully translated to a targeted RNA NGS kit that can be performed locally in decentralized sites. Since the launch of the CE-marked MP and BP NGS test, more data has been generated on this platform as well as on the established FDA-cleared microarray platform. Furthermore, decentralized sites worldwide have been onboarded and are certified to locally run the MP and BP NGS test. Methods: Paired MP and BP results were generated from FFPE RNA samples using the standard microarray as well as the MP and BP NGS test. The results from both platforms were compared to assess the concordance. Since the launch of the MP and BP NGS kit several decentralized sites underwent the onboarding process. As part of the onboarding, these sites processed a set of RNA and FFPE tissue samples previously processed at Agendia using the MP and BP NGS test. A site could only be certified if NGS results showed a 100% concordance with the Agendia results. Results: To date, over 150 RNA FFPE samples were processed with both microarray and NGS tests and MP/BP results showed concordance above 97%. Onboarding results were available for the decentralized sites. The FASTQ files generated at the sites were uploaded into the cloud-based Agendia Data Analysis Pipeline Tool (ADAPT) to generate MP and BP results. Results showed 100% concordance between Agendia’s central laboratory and the decentralized laboratories. Conclusions: MP and BP NGS test delivers equivalent results to the standard microarray test. Additionally, NGS results generated at decentralized sites also show extremely high concordance. These results confirm the high quality and robustness of the MP and BP NGS test. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Disclosure: L. Delahaye: Full / Part-time employment: Agendia. A.T. Witteveen: Full / Part-time employment: Agendia. M. Snel: Full / Part-time employment: Agendia. T. Cavness: Full / Part-time employment: Agendia. B. Chan: Full / Part-time employment: Agendia. L. Mittempergher: Full / Parttime employment: Agendia. A.M. Glas: Full / Part-time employment: Agendia.
1427P
Analysis of prognostic factors on overall survival in elderly women treated for early breast cancer using data mining and machine learning
P. Heudel1, D. Hooijenga2, R. Phan2, V. Augusto2, X. Xie2, C. Terret1, C. Faure3, S. Racadot4, O. Tredan1, T. Bachelot5 1 Department of Medical Oncology, Centre Le´on Be´rard, Lyon, France, 2CNRS, UMR 6158 LIMOS, Centre CIS, Mines Saint-Etienne, Univ Clermont Auvergne, Saint Etienne, France, 3 Surgical Oncology, Centre Le´on Be´rard, Lyon, France, 4Radiation Oncology, Centre Le´on Be´rard, Lyon, France, 5Department of Oncology, Centre Le´on Be´rard, Lyon, France Background: One third of early breast cancers (EBC) in women are diagnosed over 70 years (y) old. In this population, clinicians look not only at EBC characteristics but also antecedents and comorbidities to determine the best treatment. Methods: ConSoRe is a new generation data analytics solution using natural language processing and perform advanced data mining. It was used for data extraction from electronic patient record of 2048 patients (pts) of more than 70 years, operated for EBC in Centre Leon Berard since 1997. Patient and tumor characteristics, treatment and survival were extracted. Results: Mean age was 75.5 y (range 70-100 y). Main comorbidities described were coronary heart disease: 340 pts (16,6%) and diabetes: 321 pts (15,7%). Distribution of body mass index (BMI) was under 18.5: 56 pts (3,3%), 18.5-25: 677 pts (39,6%); 25-30: 620 (36,3%) and over 30: 356 (20,8%). Mean tumor size was 23,5 mm (range 6 to 150), SBR grading was distributed as well: grade 1: 317 (17%) ; grade 2: 1007 (55%) ; grade 3: 476 (26%) while estrogen receptor were positive (>10%) in 1553 pts (86%), progesteron receptor positive (>10%) in 1299 pts (71%) and HER2 positive in 101 pts (6,3%). Histological nodes staging was N0 in 1497 pts (72%); N1 in 414 pts (20%); N2 in 95 pts (5%); N3 in 52 pts (2%). 295 pts (14%) were treated by chemotherapy, 80 pts (3,9%) by trastuzumab, 1544 pts (75%) by radiotherapy and 1543 (75%) by hormonotherapy. Despite a limited mean follow-up of 5,1 y (range 0.3-20,7y), we observe 83 local relapse (4%), 144 metastatic relapse (7%) and 261 deaths (12.7%) with a mean overall survival (OS) of 4,6 y (10 days to 21,4 y). In the multivariate analysis, BMI under 19 and HER 2 positive were independent predictors of OS (HR: 0,85; p ¼ 2.49e-09). Using Multiple Correspondence Analysis, percentage of explained variances is very low with less than 10%. These results show the limit of classical approach by descriptive data analysis, that is why, we propose new method by machine learning and operational research to determine the best adjuvant treatment. Conclusions: This elderly population has a poor prognosis for which taking BMI into account is important when defining the therapeutic strategy. Classical approaches by data analysis reach their limits. Legal entity responsible for the study: Heudel Pierre. Funding: Has not received any funding.
Volume 30 | Supplement 5 | October 2019
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_5/mdz257.020/5576903 by guest on 24 October 2019
based on Low-Dose CT presents a series of drawbacks and needs complementary methods for improving sensitivity and specificity in the screening procedure. Thermal Liquid Biopsy (TLB) is as a complementary technique that, combined with imaging techniques, may improve the efficacy of the screening method. Methods: Blood samples from Healthy Controls (HC) and Lung Cancer Patients (LCP) were analyzed with a high sensitivity microcalorimeter VP-DSC (MicroCal – Malvern Panalytical). The data were processed in Origin 7.0 software. The plasma thermograms were analyzed through a multiparametric method developed by our research group. Statistical models allowed classifying the subjects according their serum thermograms. Results: 115 LCP subjects (average age 64.668.7, 83.0% men) with broad stage distribution (II: 5%, III: 26%; IV: 69%), smoking status (64% smoking, 7% non-smoking), histology distribution (37% adenocarcinoma, 29% squamous, 30% small cell) were compared to 119 HC subjects homogeneously distributed from a blood bank. TLB parameters obtained showed statistical differences between HC and LCP groups. Different statistical models were applied in order to establish the optimal TLB output, which is able to classify subjects according to their TLB thermogram: 92% success rate, 90% specificity, and 94% sensitivity (i.e., diagnostic odds ratio of 140). Conclusions: High positive association between clinical groups and TLB multiparametric model offers advantages over current diagnosis techniques (LDCT imaging), providing a powerful diagnostic approach with a minimally-invasive, low-risk, lowcost clinical test for LCP. Future promising applications, such as screening programs, could be developed from TLB. Legal entity responsible for the study: The authors. Funding: Instituto Carlos III (Spain). Disclosure: All authors have declared no conflicts of interest.
Annals of Oncology
1425P
Towards a screening test for cancer by circulating DNA analysis
R. Tanos1, A. Otandault1, C. Mollevi1, A. Bauer1, G. Tousch1, L. Picque Lasorsa1, S. El Messaoudi1, J. Colinge1, P-E. Colombo1, W. Jacot1, T. Mazard1, J.M. Sayague´s2, B. Gillet3, D. Pezet3, M. Ychou1, A.R. Thierry1 1 IRCM (Institute of Cancer Research of Montpellier), ICM Regional Cancer Institute of Montpellier, Montpellier, France, 2Department of Medicine and Cancer Research Center, University Hospital, Salamanca, Spain, 3Oncologie Digestive et Dermatologie, CHU Clermont-Ferrand, Clermont-Ferrand, France Background: Circulating DNA (cfDNA) has emerged as a potential biomarker in cancer, and is the subject of extensive studies in translational and clinical research. Our group has been interested in its implication and clinical significance in the field of oncology for many years, and is now focused on evaluating its potential for early cancer detection. Methods: We recently developed a screening test (MNR: Multi Normalized Ratio), based on various cfDNA parameters determined by a specific q-PCR based method, targeting both nuclear and mitochondrial sequences. Results: When applied to the supernatant of cell culture, the MNR had a discriminative potential of 100% between normal and cancer cell lines. An extensive evaluation of this test was carried out in plasma samples of 289 healthy subjects and 987 cancer patients (CRC, breast, liver, pancreatic, ovarian) of all stages. Preliminary results revealed a high potential with an AUC of 0.81 (0.78-0.84, 95% CI), a 70% sensitivity (Se) and 77% specificity (Sp). In breast cancer (N ¼ 169), an AUC of 0.82 (0.78-0.86, 95% CI) with 72% Se and 80% Sp were observed. In all stages CRC patients (N ¼ 795), the results showed an AUC of 0.80 (0.78-0.84, 95%, CI), 75% Se and 70% Sp; for CRC stages 0/I/II (N ¼ 426), an AUC of 0.79 (0.75-0.82, 95% CI), 70% Se and 72% Sp; and for CRC stage IV (N ¼ 186), a 0.86 AUC (0.82-0.89, 95% CI) with 75% Se and 80% Sp. When combining the MNR to a total cfDNA concentration threshold value (AUC ¼ 0.81 (0.790.83, 95% CI), 72% Se and 76% Sp for all stage cancers (N ¼ 987)), in a test cohort of 173 stages 0/I/II CRC patients and 132 healthy individuals, we increased the sensitivity and specificity to 74% and 95% respectively. Furthermore we recently discovered that cfDNA fragmentation, as determined by Whole Genome Sequencing using either double or single strand library, is also a parameter enabling discrimination between healthy and cancer individuals. Conclusions: The implementation of a multi-parametric test combining total cfDNA quantification, MNR and fragmentation biomarkers, with help of a decision tree in machine learning, is currently on-going. Our data suggest that our strategy in targeting cfDNA structural features might be powerful for early cancer detection, and appears as an alternative or a synergistic combination to the detection of mutations. Legal entity responsible for the study: Alain R. Thierry - INSERM (Institut national de la sante´ et de la recherche me´dicale). Funding: MSDAvenir - Mitest / Alain R. Thierry is supported by INSERM (Institut national de la sante´ et de la recherche me´dicale). Disclosure: All authors have declared no conflicts of interest.
v580 | New Diagnostic Tools
1426P
Evaluation of a successful launch of the MammaPrint and BluePrint NGS kit
L. Delahaye1, A.T. Witteveen1, M. Snel1, T. Cavness2, B. Chan3, L. Mittempergher1, A.M. Glas1 1 Product Development, Agendia Inc., Amsterdam, Netherlands, 2Operations, Agendia Inc., Irvine, CA, USA, 3Product Support, Agendia Inc., Irvine, CA, USA Background: Centralized MammaPrint (MP) and BluePrint (BP) microarray-based genomic tests on FFPE RNA were succesfully translated to a targeted RNA NGS kit that can be performed locally in decentralized sites. Since the launch of the CE-marked MP and BP NGS test, more data has been generated on this platform as well as on the established FDA-cleared microarray platform. Furthermore, decentralized sites worldwide have been onboarded and are certified to locally run the MP and BP NGS test. Methods: Paired MP and BP results were generated from FFPE RNA samples using the standard microarray as well as the MP and BP NGS test. The results from both platforms were compared to assess the concordance. Since the launch of the MP and BP NGS kit several decentralized sites underwent the onboarding process. As part of the onboarding, these sites processed a set of RNA and FFPE tissue samples previously processed at Agendia using the MP and BP NGS test. A site could only be certified if NGS results showed a 100% concordance with the Agendia results. Results: To date, over 150 RNA FFPE samples were processed with both microarray and NGS tests and MP/BP results showed concordance above 97%. Onboarding results were available for the decentralized sites. The FASTQ files generated at the sites were uploaded into the cloud-based Agendia Data Analysis Pipeline Tool (ADAPT) to generate MP and BP results. Results showed 100% concordance between Agendia’s central laboratory and the decentralized laboratories. Conclusions: MP and BP NGS test delivers equivalent results to the standard microarray test. Additionally, NGS results generated at decentralized sites also show extremely high concordance. These results confirm the high quality and robustness of the MP and BP NGS test. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Disclosure: L. Delahaye: Full / Part-time employment: Agendia. A.T. Witteveen: Full / Part-time employment: Agendia. M. Snel: Full / Part-time employment: Agendia. T. Cavness: Full / Part-time employment: Agendia. B. Chan: Full / Part-time employment: Agendia. L. Mittempergher: Full / Parttime employment: Agendia. A.M. Glas: Full / Part-time employment: Agendia.
1427P
Analysis of prognostic factors on overall survival in elderly women treated for early breast cancer using data mining and machine learning
P. Heudel1, D. Hooijenga2, R. Phan2, V. Augusto2, X. Xie2, C. Terret1, C. Faure3, S. Racadot4, O. Tredan1, T. Bachelot5 1 Department of Medical Oncology, Centre Le´on Be´rard, Lyon, France, 2CNRS, UMR 6158 LIMOS, Centre CIS, Mines Saint-Etienne, Univ Clermont Auvergne, Saint Etienne, France, 3 Surgical Oncology, Centre Le´on Be´rard, Lyon, France, 4Radiation Oncology, Centre Le´on Be´rard, Lyon, France, 5Department of Oncology, Centre Le´on Be´rard, Lyon, France Background: One third of early breast cancers (EBC) in women are diagnosed over 70 years (y) old. In this population, clinicians look not only at EBC characteristics but also antecedents and comorbidities to determine the best treatment. Methods: ConSoRe is a new generation data analytics solution using natural language processing and perform advanced data mining. It was used for data extraction from electronic patient record of 2048 patients (pts) of more than 70 years, operated for EBC in Centre Leon Berard since 1997. Patient and tumor characteristics, treatment and survival were extracted. Results: Mean age was 75.5 y (range 70-100 y). Main comorbidities described were coronary heart disease: 340 pts (16,6%) and diabetes: 321 pts (15,7%). Distribution of body mass index (BMI) was under 18.5: 56 pts (3,3%), 18.5-25: 677 pts (39,6%); 25-30: 620 (36,3%) and over 30: 356 (20,8%). Mean tumor size was 23,5 mm (range 6 to 150), SBR grading was distributed as well: grade 1: 317 (17%) ; grade 2: 1007 (55%) ; grade 3: 476 (26%) while estrogen receptor were positive (>10%) in 1553 pts (86%), progesteron receptor positive (>10%) in 1299 pts (71%) and HER2 positive in 101 pts (6,3%). Histological nodes staging was N0 in 1497 pts (72%); N1 in 414 pts (20%); N2 in 95 pts (5%); N3 in 52 pts (2%). 295 pts (14%) were treated by chemotherapy, 80 pts (3,9%) by trastuzumab, 1544 pts (75%) by radiotherapy and 1543 (75%) by hormonotherapy. Despite a limited mean follow-up of 5,1 y (range 0.3-20,7y), we observe 83 local relapse (4%), 144 metastatic relapse (7%) and 261 deaths (12.7%) with a mean overall survival (OS) of 4,6 y (10 days to 21,4 y). In the multivariate analysis, BMI under 19 and HER 2 positive were independent predictors of OS (HR: 0,85; p ¼ 2.49e-09). Using Multiple Correspondence Analysis, percentage of explained variances is very low with less than 10%. These results show the limit of classical approach by descriptive data analysis, that is why, we propose new method by machine learning and operational research to determine the best adjuvant treatment. Conclusions: This elderly population has a poor prognosis for which taking BMI into account is important when defining the therapeutic strategy. Classical approaches by data analysis reach their limits. Legal entity responsible for the study: Heudel Pierre. Funding: Has not received any funding.
Volume 30 | Supplement 5 | October 2019
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_5/mdz257.020/5576903 by guest on 24 October 2019
based on Low-Dose CT presents a series of drawbacks and needs complementary methods for improving sensitivity and specificity in the screening procedure. Thermal Liquid Biopsy (TLB) is as a complementary technique that, combined with imaging techniques, may improve the efficacy of the screening method. Methods: Blood samples from Healthy Controls (HC) and Lung Cancer Patients (LCP) were analyzed with a high sensitivity microcalorimeter VP-DSC (MicroCal – Malvern Panalytical). The data were processed in Origin 7.0 software. The plasma thermograms were analyzed through a multiparametric method developed by our research group. Statistical models allowed classifying the subjects according their serum thermograms. Results: 115 LCP subjects (average age 64.668.7, 83.0% men) with broad stage distribution (II: 5%, III: 26%; IV: 69%), smoking status (64% smoking, 7% non-smoking), histology distribution (37% adenocarcinoma, 29% squamous, 30% small cell) were compared to 119 HC subjects homogeneously distributed from a blood bank. TLB parameters obtained showed statistical differences between HC and LCP groups. Different statistical models were applied in order to establish the optimal TLB output, which is able to classify subjects according to their TLB thermogram: 92% success rate, 90% specificity, and 94% sensitivity (i.e., diagnostic odds ratio of 140). Conclusions: High positive association between clinical groups and TLB multiparametric model offers advantages over current diagnosis techniques (LDCT imaging), providing a powerful diagnostic approach with a minimally-invasive, low-risk, lowcost clinical test for LCP. Future promising applications, such as screening programs, could be developed from TLB. Legal entity responsible for the study: The authors. Funding: Instituto Carlos III (Spain). Disclosure: All authors have declared no conflicts of interest.
Annals of Oncology