Deliverable

8.2Scenario E-Retrospective use of clinical data

Scenario E
Objective	An academic researcher wants to define if pCR is a candidate surrogate marker for Disease Free Survival (DFS) and Overall Survival (OS) independently of treatment type.
Steps	The researcher logs into the system. The researcher filters by type of cancer (i.e. breast), the treatment setting (i.e. neoadjuvant), selected treatment (i.e. all) and the pathological characteristics (i.e. all). The researcher selects the outcome data (i.e. pCR, DFS, OS). The researcher either downloads the results on his computer (i.e. an excel file in csv format) or works directly in the INTEGRATE platform using the provided tools and defines how pCR correlates to DFS and OS independently of treatment type.
Results	According to the obtained results, the academic researcher will design a new NeoBIG trial in which pCR will or will not be used as a surrogate endpoint.

Table Scenario E-Retrospective use of clinical data

In this scenario, a researcher wants to investigate the association between the pathological complete response (pCR) and clinical outcome in terms of Disease Free Survival (DFS) and Overall Survival (OS) independently of any treatment type. According to the obtained results, the researcher will design a new NeoBIG trial in which pCR will or will not be used as a surrogate endpoint.

Initially, the researcher logs into the INTEGRATE platform and exports the examined dataset which consists of patients with breast cancer, treated by any possible type of regimen under neoadjuvant therapy. The examined dataset is dichotomized into two groups using the binary variable pCR as the independent variable. In complex diseases, such as cancer, researchers rely on statistical comparisons of DFS and OS of patients against healthy control groups or against patients following different treatment as in [52]. In this approach DFS and OS will be estimated by Kaplan-Meier survival analysis [53] and the log-rank test will be used to compare DFS and OS between the two groups (pCR VS no pCR achieved).
In this statistical analysis, Kaplan-Meier survival curves, along with the 95% confidence interval for the curves, showing the DFS and OS of pCR and non-pCR patients treated by any type of neoadjuvant regimen will be presented. A representative example, given by [54], is illustrated below. Anthracycline-treated patients were dichotomized by their Tissue Inhibitor of Metalloproteinases-1 (TIMP-1) level and Kaplan-Meier survival analysis has been done, showing the cumulative percentages of DFS (subfigure A) and OS (subfigure B) over time.

Figure Kaplan-Meier plot showing the DFS (A) and OS (B) of TIMP-1 [54].

To compare the survival distributions given by our groups, the widely-used non-parametric log-rank test will be performed. It provides a p value that indicates whether or not the difference in survival between the two groups is statistically significant. Therefore, estimating the log-rank between the survival curve of pCR and non-pCR groups we interpret a p value that indicates a statistically significant deference (low p values) or a convergence of the two curves if the p value is high.
After the completion of the statistical analysis, the researcher either works directly to the INTEGRATE platform or downloads all the analysis to his/her local computer. The downloaded analysis could be an excel file with the resulted tables and graphical results placed in the same sheet.

8.3Scenario F-Retrospective use of imaging data

Scenario F
Objective	An academic researcher wants to define if pCR is associated with a decrease of more than 20% in tumor volume between baseline and day 15, so that decrease in tumor volume between baseline and day 15 could be used as an early surrogate for pCR.
Steps	The researcher logs into the system. The researcher filters by type of cancer (i.e. breast) and the treatment setting (i.e. neoadjuvant). The researcher selects for these outputs: the imaging data at baseline and at day 15 and response to treatment. The researcher either downloads the results on his computer (i.e. an excel file in csv format) or works directly in the INTEGRATE platform using the provided tools and defines if pCR is associated with a decrease of more than 20% in tumor volume between baseline and day 15 or not.
Results	According to the obtained results, the researcher will design a new NeoBIG trial to validate if the decrease in tumor volume between baseline and day 15 could be used as an early surrogate marker for pCR.

Table Scenario F-Retrospective use of imaging data

Tumor volume can be extracted from the information contained in the tags of

standard DICOM images (MRI, CT etc.) and the delineation/segmentation of
the tumor by the doctors, using the "DrEye" tool by FORTH [55]. DICOM stands for Digital Imaging and Communications in Medicine, and it is a standard for handling, storing, printing, and transmitting information in medical imaging. It includes a file format definition and a network communications protocol. The National Electrical Manufacturers Association (NEMA) holds the copyright to this standard. It was developed by the DICOM Standards Committee, whose members are also partly members of NEMA.

The DICOM format groups information into data sets. That means that a file

of a chest X-Ray image, for example, actually contains the patient ID within
the file, so that the image can never be separated from this information by
mistake. This is similar to the way that image formats such as JPEG can also
have embedded tags to identify and otherwise describe the image.
A DICOM data object consists of a number of attributes, including items such
as name, ID, etc., and also one special attribute containing the image pixel
data (i.e. logically, the main object has no "header" as such: merely a list
of attributes, including the pixel data). A single DICOM object can only
contain one attribute containing pixel data. For many modalities, this
corresponds to a single image. But note that the attribute may contain
multiple "frames", allowing storage of cine loops or other multi-frame data.
Another example is NM data, where an NM image by definition is a
multi-dimensional multi-frame image. In these cases three- or
four-dimensional data can be encapsulated in a single DICOM object. Pixel
data can be compressed using a variety of standards, including JPEG, JPEG
Lossless, JPEG 2000, and Run-length encoding (RLE). LZW (zip) compression
can be used for the whole data set (not just the pixel data) but this is
rarely implemented.
The estimation of the tumor’s volume is based on the fact that the
tumor is a collection of voxels. A voxel can be defined as the volume unit,
which can be computed from the information contained in the DICOM tags of
each image in the series of interest. The volume of a voxel is the product
of the information in the tag by the sum of the information in
the tag and in tag. These calculations are given by the following equation:

With the volume unit (=voxel) to be defined, a draft estimation of volume of
the tumor is computed based by multiplying the sum of all the voxels of the
tumor with the volume unit. The estimation of the volume can be improved by
interpolating the available series and creating a new series in an isotropic
space where the volume unit is defined as 1 cubic mm, which is
fairly smaller than the one without interpolation.
Having the available information from the DICOM imaging system, we estimate the tumor volume at baseline and day 15 for each patient. All cases are dichotomised based on if they achieve to perform a decrease in volume change equal or more than 20%. A confusion matrix is then given between the pCR and the decrease of tumor change, as depicted in Table . An odds ratio statistical analysis will be performed to characterize the association between the two variables and validate if the decrease in tumor volume between baseline and day 15 could be used as an early surrogate marker for pCR. A graphical representation of the odds ratios through forest plots will be also provided, as depicted in Figure .

	20% decrease	20% decrease
pCR	a	b
No pCR	c	d

Table Confusion matrix for tumor volume change

9REFERENCES

1. 
   P. Carmeliet, R. K. Jain, “Angiogenesis in cancer and other diseases”, Nature, vol. 407 pp. 249-259, 2002.
   
2. 
   Cristini, V., Frieboes, H.B., Gatenby, R., Caserta, S., Ferrari, M., Sinek, J.P. Morphological instability and cancer invasion. Clin. Cancer Res. 11, 6772–6779, 2005.
   
3. 
   http://en.wikipedia.org/wiki/Virtual_Physiological_Human
   
4. 
   Fenner JW, Brook B, Clapworthy G, Coveney PV, Feipel V, Gregersen H, Hose DR, Kohl P, Lawford P, McCormack KM, Pinney D, Thomas SR, Van Sint Jan S, Waters S, Viceconti M. The EuroPhysiome, STEP and a roadmap for the virtual physiological human. Philos Transact A Math Phys Eng Sci. 366(1878):2979-99, 2008.
   
5. 
   L. B. Edelman, J. A. Eddy, and N. D. Price, In silico models of cancer. Wiley & Sons, 2009.
   
6. 
   Dedeurwaerder S, Desmedt C, Calonne E, Singha SK, Haibe-Kains B, Defrance M, Michiels S, Volkmar M, Deplus R, Luciani J, Lallemand F, Larsimont D, Toussaint J, Haussy S, Rothé F, Rouas G, Metzger O, Majjaj S, Saini K, Putmans P, Hames G, Baren NV, Coulie PG, Piccart M, Sotiriou C, Fuks F. DNA methylation profiling reveals a predominant immune component in breast cancers. EMBO Mol Med. 2011.
   
7. 
   Uwe Scherf et al., A gene expression database for the molecular pharmacology of cancer, nature genetics, volume 24, March 2000.
   
8. 
   G. Stamatakos, D. Dionysiou, N. Mouravliansky, K. Nikita, G. Pissakas, P. Georgolopoulou and N. Uuznoglu, “Algorithmic Description of the Biological Activity of a Solid Tumor in Vivo”, in Proc. EUROSIM 2001 Congress, Delft, the Netherlands, June 26-29, 2001 (CD-ROM Edition).
   
9. 
   K.Swanson, E.C.Alvord Jr., J.D.Murray, “Dynamics of a model for brain tumors reveals a small window for the therapeutic intervention,” Discrete and Continuous Dynamical Systems-Series B, vol. 4, no 1, pp.289-295, 2004.
   
10. 
    Olivier Glatz, Maxime Sermesant, Pierre-Yves Bondiau, Hervé Delingette, Simon K. Warfield, Grégoire Malandain, and Nicholas Ayache. Realistic Simulation of the 3D Growth of Brain Tumors in MR Images Coupling Diffusion with Mass Effect. IEEE Transact. on Medical Imaging, 24(10):1334-1346, 2005.
    
11. 
    Benjamin Ribba, Thierry Colin and Santiago Schnell, A multiscale mathematical model of cancer, and its use in analyzing irradiation therapies, Theoretical Biology and Medical Modelling, 3:7, 2006.
    
12. 
    Graham A. Colditz and A. Lindsay Frazier, Models of Breast Cancer Show That Risk Is Set by Events of Early Life: Prevention Efforts Must Shift Focus, Cancer Epidemiology, Biomarkers & Prevention, Vol. 4. 567-571. July/August 1995.
    
13. 
    Schmid P, Wischnewsky MB, Sezer O, Böhm R, Possinger K:Prediction of Response to Hormonal Treatment in Metastatic Breast Cancer. Oncology; 63:309-316, 2002.
    
14. 
    Wolmark N, Wang J, Mamounas E, Bryant J, Fisher B. Preoperative chemotherapy in patients with operable breast cancer: nine-year results from National Surgical Adjuvant Breast and Bowel Project B-18. Journal of the National Cancer Institute, (30):96-102, 2001.
    
15. 
    C Desmedt, A Di Leo, E de Azambuja, D Larsimont, B Haibe-Kains, J Selleslags, S Delaloge, C Duhem, J-P Kains, B Carly, M Maerevoet, A Vindevoghel, G Rouas, F Lallemand, V Durbecq, F Cardoso, R Salgado, R Rovere, G Bontempi, S Michiels, M Buyse, J-M Nogaret, Y Qi, F Symmans, L Pusztai, V D'Hondt, M Piccart-Gebhart and C Sotiriou. Multifactorial Approach to Predicting Resistance to Anthracyclines. American Society of Clinical Oncology, Volume 29, Number 12, April 2011.
    
16. 
    R. Somorjai, B. Dolenko, and R. Baumgartner. Class prediction and discovery using gene microarray and proteomics mass spectroscopy data: curses, caveats, cautions. Bioinformatics, 19(12):1484–1491, 2003.
    
17. 
    H. Liu and L. Yu. Toward integrating feature selection algorithms for classification and clustering. IEEE Transactions on Knowledge and Data Engineering, 17(4):491–502, 2005.
    
18. 
    Saeys, Y., Inza, I. & Larrañaga, P. A review of feature selection techniques in bioinformatics. Bioinformatics 23, 2507-2517, 2007.
    
19. 
    Yee Hwa Yang, Yuanyuan Xiao, and Mark R. Segal. Identifying differentially expressed genes from microarray experiments via statistic synthesis. Bioinformatics 21, 1084-1093, April 2005.
    
20. 
    S. Yu, T. Falck, A. Daemen, L.-C. C. Tranchevent, J. A. A. Suykens, B. De Moor, and Y. Moreau. L2-norm multiple kernel learning and its application to biomedical data fusion. BMC bioinformatics, vol. 11, no. 1, pp. 309+, Jun. 2010.
    
21. 
    Wilcoxon, Frank. Individual comparisons by ranking methods. Biometrics Bulletin 1 (6): 80–83, Dec. 1945.
    
22. 
    J. W. Lee, J. B. Lee, M. Park, and S. H. Song. An extensive comparison of recent classification tools applied to microarray data. Comp. Statistics and Data Analysis, vol. 48, pp. 869–885, 2005.
    
23. 
    Sutton, C.D. Classification and Regression Trees, Bagging, and Boosting. Handbook of Statistics 24, 303–329, 2005.
    
24. 
    Gevaert, O. A Bayesian network integration framework for modeling biomedical data. Ph.D dissertation, Katholieke Universiteit Leuven, 2008.
    
25. 
    Vapnik, V. The Nature of Statistical Learning Theory. Springer, N.Y. 1995.
    
26. 
    B. Scholkopf, A. J. Smola. Learning with Kernels. The MIT Press, 2002.
    
27. 
    Mika, S., Ratsch, G., Weston, J., Scholkopf, B., Smola, A.J., Mueller, K.-R. Constructing descriptive and discriminative non-linear features: Rayleigh coefficients in kernel feature spaces. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004.
    
28. 
    J. Shawe-Taylor and N. Cristianini. Kernel Methods for Pattern Analysis. Cambridge University Press, 2004.
    
29. 
    G. Lanckriet, N. Cristianini, P. Bartlett, L. El Ghaoui, and M. Jordan. Learning the kernel matrix with semi-definite programming. Journal of Machine Learning Research, 5, 2004.
    
30. 
    P. Pavlidis, J. Cai, J. Weston and W. Noble Grundy. Gene functional classification from heterogeneous data. In Proceedings of the Fifth Annual International Conference on Computational Biology: April 22-25, 2001.
    
31. 
    Pai-Hsuen Chen, Chih-Jen Lin and Bernhard Scholkopf, “A Tutorial on ν-Support Vector Machines”. Applied Stochastic Models in Business and Industry, Vol. 21, pp. 111-136, No. 2. 2005.
    
32. 
    A. J. Smola, P. L. Bartlett, B. Schölkopf, and D. Schuurmans. Advances in Large Margin Classifiers. MIT Press, Cambridge, MA, 2000.
    
33. 
    Breiman, L., Bagging predictors. Machine Learning 24 (2), 123–140, 1996.
    
34. 
    Freund, Y., Schapire, R.E., Experiments with a new boosting algorithm. In Proceedings 13th International Conference on Machine Learning. Morgan Kaufmann, Bari, Italy, pp. 148–156, 1996.
    
35. 
    Breiman, L., Random forests. Machine Learning 45 (1), 5–32, 2001.
    
36. 
    Rodríguez, J.J., Kuncheva, L.I., Alonso, C.J. Rotation forest: A new classifier ensemble method. IEEE Trans. Pattern Anal. Machine Intell. 28 (10), 1619–1630, 2006.
    
37. 
    Meir, R., Rätsch, G. An introduction to boosting and leveraging. In: Advanced Lectures on Machine Learning. Lecture Notes Comput. Sci., vol. 2600. Springer-Verlag, Berlin, pp. 118–183, 2003.
    
38. 
    Jin, R., Zhang, J. Multi-class learning by smoothed boosting. Machine Learning. 67 (3), 207–227, 2007.
    
39. 
    Quinlan, J. R. C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, 1993.
    
40. 
    Smeeton, N.C. Early History of the Kappa Statistic. Biometrics 41: 795, 1985.
    
41. 
    Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 33:159–174, 1977.
    
42. 
    David, A., Comparison of classification accuracy using Cohen’s weighted kappa. Expert Syst. Appl. 825-832, 2008.
    
43. 
    Fawcett, T. An introduction to ROC analysis, Pattern Recognition Letters 27(8),861-874, 2006.
    
44. 
    Z. Chen, J. Li, and L. Wei. A multiple kernel support vector machine scheme for feature extraction and rule extraction from gene expression data of cancer tissue. Artificial Intelligence in Medicine, vol. 41, no. 2, pp. 161-175, October 2007.
    
45. 
    M. Varma and D. Ray. Learning the discriminative power-invariance trade-off. Proceedings of the IEEE International Conference on Computer Vision, pp. 1-8, October 2007.
    
46. 
    A. Zien and C. S. Ong. Multiclass multiple kernel learning. Proceedings of the 24th International Conference on Machine Learning, pp. 1191-1198, October 2007.
    
47. 
    Ye, J., Alexander, G., Wu, T., Chen, K., Wu, T., Li, J., Zhao, Z. Heterogeneous data fusion for Alzheimer’s disease study. Proceeding of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining KDD 08, 2008.
    
48. 
    Ran Gilad-Bachrach, Amir Navot and Naftali Tishby. Margin Based Feature Selection-Theory and Algorithms. Proceedings of the 21st International Conference on Machine Learning, Canada, p. 43-50, 2004.
    
49. 
    K.Kira and L.Rendell. A practical approach to feature selection. Proceedings of 9th International Workshop on Machine Learning, p 249-256, 1992.
    
50. 
    S. Lewis and M. Clarke. Forest plots: trying to see the wood and the trees. BMJ (Clinical research ed.), 322(7300):1479–1480, June 2001.
    
51. 
    http://en.wikipedia.org/wiki/Forest_plot
    
52. 
    S. Kaura and G. Dranitsaris. Letrozole or anastrozole for the treatment of hormone-positive breast cancer: A clinical comparison using indirect statistical techniques. American Society of Clinical Oncology, Volume 28, Number 15, May 2010.
    
53. 
    Kaplan, E.L., Meier, P. Nonparametric estimation from incomplete observations. Journal of the American Statistical Association, 53, 457-481 1958.
    
54. 
    Schrohl AS, Look MP, Meijer-van Gelder ME, Foekens JA, Brünner N. Tumor tissue levels of Tissue Inhibitor of Metalloproteinases-1 (TIMP-1) and outcome following adjuvant chemotherapy in premenopausal lymph node-positive breast cancer patients: A retrospective study. BMC Cancer. 10; 9:322, Sep. 2009.
    
55. 
    http://biomodeling.ics.forth.gr