RERTEST. MSI-H PREDICTIVE MODEL IN COLORECTAL CARCINOMA
Microsatellite instability (MSI-H) in colorectal carcinoma is a molecular parameter of clinical importance due to its prediction and prognostic values as well as for its relation with a high incidence of Lynch syndrome (HNPCC). RERTEST6 is a model that allows the prediction of MSI-H in colorectal carcinoma with a negative prediction value of 97,9%, using six parameters looked at in the routine pathological examination. A second version, RERTEST8, also incorporates two immunohistochemical parameters, the percentages of expression of Ki67 and p53. These models have been elaborated in the last years from a project conceived in the group of Biopat and Histopat, in collaboration with the department of Statistics in the Faculty of Biology, University of Barcelona and with the Pathology department of different hospitals from Barcelona and province (Hospital de Barcelona-SCIAS, Hospital Universitari de Bellvitge, Hospital del Mar, Hospital Mútua de Terrassa, Hospital General de Vic and Hospital Plató). The work has included a total of 615 cases of primary colorectal adenocarcinoma with a histopathologic examination evaluating 22 parameters, immunohistochemistry (Ki67, p57) and the molecular study of microsatellite instability. The initial statistical analysis was based on a classical process of selection known as “stepwise regression”. The preliminary results have been presented in different national and international congresses and were initially published in North America (Diagn Mol Pathol 2005; 14: 213-223). The definite results, after a multicentric validation and optimization study, have been recently published in Europe (Virchows Arch 2010; 456: 533-541). A new methodology “Regularization Paths for Generalized Linear Models via Coordinate Descent” described by Friedman has been used in the optimization study. Biopat makes available both prediction models in an “Excel” format, trusting that their incorporation in the routine pathological study can provide complementary information with clinical utility. We remain at your disposal for any additional information that may be required (firstname.lastname@example.org).