Profili di espressione genica Giampaolo Bianchini MD Ospedale San Raffaele, Milan - Italy
Gene expression profiles Transcriptomics Gene DNA mrna mirnas Protein metilation Metabolite Genomics Transcriptomics Proteomics Epigenomics Regulomics Metabolomics CGH arrays SNP arrays Whole genome Sequencing Methyl-DIP Two channel arrays Single channel arrays RNA Microarray sequencing SELDI-TOF mass spectrometry Mass spectrometry
mrna expression Wide dynamic range
Protein (IHC) versus mrna expression ER- (0%) No benefit ER+ (50%) ER+ (100%) Lower benefit Higher benefit ER- ER+ Callari M Clin Cancer Res 2016
Percentage of ER protein expression (by IHC) is neither prognostic nor predictive in endocrine treated patients ER-poor (ER < 80%) ER-rich (ER 80%) Technical limitations of ER assessment by IHC: Small dynamic range Skewed distribution (>75% of the tumors 90%-100% staining) Viale G Annals Oncol 2011
ER expression by RT-PCR (Oncotype DX) is predictive of tamoxifen benefit (NSABP-14 trial) High Tertiles Intermediate Tertiles Low Tertiles Kim C JCO 2011
Wide-dynamic range and normal distribution of ER expression by RT-PCR Paik S SABCS 2004
mrna expression Technical/Assay validity High accuracy
Inter-observer variability
Technical/Assay validity of Ki67 is limited Conclusions Substantial variability in Ki67 scoring was observed among some of the world s most experienced laboratories. Ki67 values and cutoffs for clinical decisionmaking cannot be transferred between laboratories without standardizing scoring methodology because analytical validity is limited. Mei-Yin C JNCI 2013
Technical/Assay validity for multigene assay has been provided Publications that provide information on analytical validity i.e. Oncotype DX (Azim et al Ann Oncol 2013;24:647) High reproducibility for repeat analyses across multiple dates, PCR plates and instruments (SD in RS < 1) High repeatability from different (2-5) blocks from same patient (between block SD in RS < 2.5) High repeatability within FFPE block (within block SD in RS < 1) Failure rates (usually less than 5%)
Gene expression profiles Endowed of widest possible applicability (suitable for FFPE tissue)
Which level of evidence is provided for Oncotype DX? Prospective (randomized) clinical trials to evaluate the medical utility of a prognostic or predictive biomarker are the gold standard. However, such trials are costly and require long period of time New guidelines has been suggested for indirect prospectiveretrospective designs using archived specimens Simon R, Paik S Hayes DF. JNCI 2009
Gene expression profiles Answer a therapeutically relevant question
Answer a therapeutically relevant question Intended use of the biomarker Validation = Fitness for Intended Use (Simon R.)
Relevant therapeutic question in ER+/HER2- tumors Can we tailor adjuvant chemotherapy administration? NSABP B-20 trial CMF/TAM TAM These patients could be safely spared by CT
Several mutigene assays have been developed for tailoring treatment in ER+/HER2- patients
Recurrence Score (RS, Oncotype DX) development Panel of 21 Genes and the Recurrence - Score Algorithm Training set 250 genes RT - PCR in FFPET 447 patients Three studies Paik S NEJM 2004
Oncotype DX: First validation study (NSABP-14 trial) ER+, node negative, tamoxifen-treated patients Risk Category Percentage of Patients Rate of Distant Recurrence at 10 Yr (95% CI) Rate of Death at 10 Yr (95% CI) Low 51 6.8 (4.0-9.6) 3.1 (1.2-5.0) Intermediate 22 14.3 (8.3-20.3) 12.2 (6.7-17.6) Medical utility requires an actionable prediction to be therapeutically relevant High 27 30.5 (23.6-37.4) 27.0 (20.4-33.6) Paik S et al NEJM 2004
Second validation study (ATAC trial) ER+, node negative, endocrine treated patients Prediction performances independently confirmed n=872 Dowsett M J Clin Oncol 2010
Oncotype DX in patients with RS<11
Gene expression profiles Suitable to be used as continuous markers
Do we need dicotomic risk group for dicotomic clinical decisions? Molecular assay Low risk High risk NO Chemo Chemo
Dichotomized biological variables can lead to lose of information
Biology is a continuum Risk is a continuum NO Chemo Chemo Dicotomic clinical decision based on continuous risk assessment allowed for a more tailored decidion making Dicotomic markers are significantly less informative Chemo RS=18 NO Chemo RS=17 Paik S NEJM 2004
Gene expression profiles Study design to validate predictive markers
Validation of predictive markers for chemotherapy benefit using prospective-retrospective studies % HIGH % 100 50 CT Endocrine 100 Is this result enough CT Endocrine to claim that the test is predictive of 50 chemotherapy benefit? Endocrine 0 100 % Endocrine HR 0.65 (0.50-0.90) p=0.009 0 5 10 15 CT Endocrine 0 50 0 5 10 15 NO! LOW 0 Endocrine HR 0.95 (0.80-1.15) p=0.88 0 5 10 15
Validation of predictive markers for chemotherapy benefit using prospective-retrospective studies % 100 the different treatment benefit that you see in CT the Endocrine Y high and low 0 5 10 15 HIGH You need to demonstrate that This information is provided 0 by a test for % 0 5 10 15 LOW % 100 50 50 0 CT Endocrine Endocrine HR 0.65 (0.50-0.90) p=0.009 50 group cannot be explain by interaction between treatment benefit and chance only (that the benefit is 100 Endocrine biomarker group statistically different by 0 treatment group) CT Endocrine Endocrine HR 0.95 (0.80-1.15) p=0.88 0 5 10 15
Predictive value of chemotherapy (CMF) benefit by Recurrence Score (hypothesis generation retrospective-prospective study) NSABP B-20 trial (ER+, node negative) Tamoxifen versus CMF/TAM RS<18 RS 18-30 Paik S JCO 2006 RS 31
Predictive value of chemotherapy (CMF) benefit by Recurrence Score (independent retrospective-prospective study) SWOG-8814 trial (ER+, node positive) Tamoxifen versus CAF TAM RS<18 RS 18-30 Albain KS Lancet Oncol 2010 RS 31
Prospective randomized clinical trials are ongoing to provide the level of evidence I for the predictive value of Oncotype DX TAILORx trial RxPONDER trial ER+/HER2-, node negative ER+/HER2-, node positive (1-3) Oncotype DX assay Oncotype DX assay RS 11-25 RS 25 R R Chemotherapy and hormonal treatment Hormonal treatment only Chemotherapy and hormonal treatment Hormonal treatment only
Predictive value of Mammaprint Cardoso F NEJM 2016
Gene expression profiles Combining markers
Is it about multi-gene signatures against pathologists? Multi-gene signatures Pathologist
Anatomy and Biology: two complementary sides of breast cancer prognostication Should we integrate molecular and clinico-pathological information? Tang G et al. J Clin Oncol 2011
RSPC tool has a user friendly interface Gong G JCO 2011
RSPC (Recurrence Score-Pathology-Clinical) Aromatase inhibitor Age 60 Grade 2 pt 1 cm 10 yrs RISK 7% (5%-10%) RS=23 10 yrs RISK 12% (9%-16%) Average risk Average pathology-clinical characteristics Age 60 Grade 3 pt 3 cm 10 yrs RISK 27% (20%-36%) Gong G JCO 2011
Patients with very good or very poor prognostic features might not need to be tested Age 50 Grade 1 pt 1.5 cm RS=11 4% (3%-6%) Utility of gene expression profiles is where RS=17 clinical uncertainty 6% (4%-9%) is present RS=25 8% (5%-12%) RS=35 10% (6%-15%)