Table 9Evidence Table. Clinical utility and validity, MammaPrint®

Study, yearMeasureConclusions
Buyse, 200659Context: This study was designed to test the validation of the MammaPrint assay through the TRANSBIG consortium multicenter, multinational collaborative study involving 5 distinct institutions.MammaPrint is a better predictor of TTM than Age, Size, Grade, ER, Adjuvant!, NPI, St Gallen;
End points: OS, RFS (any recurrence as event); TTM (time to metastases as first relapse)St Gallen is a better predictor of DFS than MammaPrint
Methods:
  • The Kaplan-Meier product limit estimator was used in survival analyses.
  • Hazard ratios were formulated to compare the high-risk group with the low-risk group.
  • Hazard ratios heterogeneity across institutions was evaluated by chi-square test with four degrees of freedom.
  • Gene signature hazard ratios were stratified by clinical risk and vice versa.
  • Follow-up duration impact on hazard ratios was analyzed by censoring all observations at increasing time points.
  • Sensitivity and specificity were estimated for distant metastases within 5 years and for death within 10 years.
  • Time-dependent receiver operating characteristic (ROC) curves were computed by using the tumor expression level for the 70-genes signature and the 10-year survival probability for the Adjuvant! Software.
MammaPrint is a better predictor for OS than Age, Size, Grade, ER, Adjuvant!, NPI, St Gallen;
Results:
  • Kaplan-Maier analysis stratified by MammaPrint and Adjuvant (% of patients with distant recurrence):
The signature remained a statistically significant prognostic factor for TTM and OS even after adjustment for various risk classifications methods based on clinicopathologic factors;
 Good(R>0.4), Adjuvant!Low: 52 patientsThe lack of statistical significance for DFS was explained by the fact that the signature was originally developed using TTM as the endpoint.
  OS(10years): 0.88 (0.74 to 0.95)Overall the 70-gene signature adds independent prognostic information to clinicopathologic risk assessment for node-negative untreated patients with early breast cancer;
 Good(R>0.4)Adjuvant!High: 59 patientsClinical risk hazard ratios, adjusted for the gene signature were not significant, suggesting that most of their prognostic utility is subsumed by the gene signature;
  OS(10years): 0.89 (0.77 to 0.95)
 Poor(R<0.4)Adjuvant!Low: 28 patients
  OS(10years): 0.69 (0.45 to 0.84)
 Poor(R<0.4)Adjuvant!High: 163 patients
  OS(10years): 0.69 (0.61 to 0.76)
  • Hazard Ratios (unadjusted):
 MammaPrint:
  TTM=2.32 (95% CI = 1.35–4.00)
  DFS=1.50 (95% CI = 1.04–2.16)
  OS=2.79 (95% CI = 1.60–4.87)
 Adjuvant:
  TTM=1.68 (95% CI = 95% CI = 0.92 to 3.07)
  DFS=1.30 (95% CI = 95% CI = 0.86 to 1.95);
  OS=1.67 (95% CI = 95% CI =0.93 to 2.98)
  • Hazard Ratios (adjusted):
 MammaPrint adjusted by St Gallen:
  TTM=2.15 (95% CI = 1.25 to 3.71)
  DFS=1.41 (95% CI = 0.97 to 2.06)
  OS=2.69 (95% CI = 1.53 to 4.73)
 MammaPrint adjusted by NPI:
  TTM=2.15 (95% CI =1.19 to 3.92)
  DFS=1.45 (95% CI = 0.97 to 2.16)
  OS=2.89 (95% CI = 1.58 to 5.29)
 MammaPrint adjusted by Adjuvant:
  TTM= 2.13 (95% CI = 1.19 to 3.82)
  DFS= 1.36 (95% CI = 0.91 to 2.03)
  OS= 2.63 (95% CI =1.45 to 4.79)
 Adjuvant Adjusted by MammaPrint:
  TTM=1.26 (95% CI = 0.66 to 2.40);
  DFS=1.03 (95% CI = 0.67 to 1.59)
  OS=1.08 (95% CI = 0.59 to 1.99)
  • Sensitivity for development of metastases within 5 years:
 Gene signature 0.90 (0.78 to 0.95)
 Adjuvant! software 0.87 (0.75 to 0.94)
 Nottingham Prognostic Index 0.79 (0.65 to 0.88)
 St Gallen criteria 0.96 (0.86 to 0.99)
  • Specificity for development of metastases within 5 years:
 Gene signature 0.42 (0.36 to 0.48)
 Adjuvant! software 0.29 (0.24 to 0.35)
 Nottingham Prognostic Index 0.48 (0.42 to 0.54)
 St Gallen criteria 0.10 (0.07 to 0.14)
  • Sensitivity for death within 10 years:
 Gene signature 0.84 (0.73 to 0.92)
 Adjuvant! software 0.82 (0.71 to 0.90)
 Nottingham Prognostic Index 0.70 (0.57 to 0.80)
 St Gallen criteria 0.95 (0.86 to 0.98)
  • Specificity for death within 10 years:
 Gene signature 0.42 (0.36 to 0.48)
 Adjuvant! software 0.29 (0.23 to 0.35)
 Nottingham Prognostic Index 0.47 (0.41 to 0.53)
 St Gallen criteria 0.10 (0.07 to 0.14)
  • ROC area under the curve for MammaPrint:
 TTM: 0.681
 OS: 0.659
  • ROC area under the curve for Adjuvant:
 TTM: 0.648
 OS: 0.576
  • Adjustments for clinical risk using different cutoff points produced only minor changes in the adjusted gene signature hazard ratios
Espinosa, 200580Context: 70-genes signature partial validation using RT-PCR analysis: 60 out of the 70 genes of the signatures plus 4 additional genes. The study was conducted to reproduce the results obtained by microarray by an alternative method using 96 patients with full follow-up and frozen specimen from a single institutionqRT-PCR reproduced the results obtained with microarrays for the 70-genes prognostic profile in women with early-stage breast cancer
End points: Relapse-free survival (RFS), and overall survival (OS) measured by relapses dichotomization (regardless of the site of first relapse): low-risk if the patient survived 2 years or longer, or high risk otherwiseBoth RFS and OS were significantly longer in patients harboring a good-prognosis gene profile.
Methods:
  • The RT-PCR based 70-genes signature surrogate: was obtained as follows:
 Gene by gene qRT-PCR normalization,
 Gene expression categorization using the cutoff value of 0,
 Distance/similarity coefficient calculation for each patient,
 Similarity cutoff value optimization to yield the highest significance in log-rank tests for DFS
  • The Kaplan-Meier product limit method was used in survival analysis and significance was assessed by bilateral log-rank test
  • The Mann-Whitney test was used to compare patients'; characteristics
  • The Fisher's exact test was used to compare categorical variables
  • Multivariate Cox proportional regression analysis was performed by adjusting by ER status, lymph node status, number of positive lymph nodes (less or more than 3) and gene profile
Results:
  • Kaplan-Meier estimates for the Good prognosis profile:
 RFS after 70 months: 85%
 OS after 70 months: 97%
 Log-rank test P value = 0.03
  • Kaplan-Meier estimates for the Bad prognosis profile:
 RFS after 70 months: 62%
 OS after 70 months: 72%
 Log-rank test P value = 0.002
  • Results of multivariate Cox proportional regression analysis:
 OS by lymph node status, HR = 1.2; 95%CI = 1.09 to 1.36
 OS by the 70-genes profile HR = 6.3; 95%CI = 1.28 to 31.07
 DFS by lymph node number, HR = 1.13; 95%CI = 1.05 to 1.25
 DFS by the 70-genes profile HR = 2.74, 95%CI = 1.13 to 6.61
Fan, 200679Context: The study is a comparison of classification agreement among 5 different gene expression based predictors (70-genes, Oncotype DX™, Wound-response, Intrinsic subtypes and two-gene ratio) using the 295 samples from the consecutive cohort originally used by van de Vijver 200225.Good but not perfect correlation between predictions, albeit surprising with different gene sets. Degree of prediction over and above “standard” clinical stratifiers is not clear - reclassification not done.
Methods:
  • Cramer V statistics
  • Kaplan-Meier survival analysis
  • Multivariate Cox proportional hazards analysis, adjusting for age, tumor size, tumor grade, ER status, number of lymph node involved
  • All analyses were performed using all the patients, as well as the ER positive patients subset (N=225)
  • NB: coefficients of clinical predictors were allowed to vary between models
Results:
  • In multivariate Cox proportional hazards analysis all tests except the two-gene ratio were highly significant predictors of OS and DFS:
 70 genes DFS HR = 3.4 (95%CI = 2.0–6.0), P value < 0.001
 Two-gene ratio DFS HR = 0.91 (95%CI = 0.6–1.3), P value = 0.62
 Oncotype OS HR = 4.3 (95%CI = 2.1–8.9), P value < 0.001
 70 genes OS HR = 4.71 (95%CI = 2.02–11.0), P value < 0.001
 Two-gene ratio OS HR = 1.00 (95%CI = 0.61–1.63), P value = 0.99
 Oncotype DFS HR = 6.14 (95%CI = 1.84–20.4), P value = 0.003
  • ER status, tumor grade, tumor size, and lymph nodes were also significant predictors.
  • Patients classification into dichotomized risks groups proved to be roughly similar, ranging from ~80% to ~40%:
 70-genes and RS yielded a V= 0.60
 70-genes and RS yielded an agreement of ~81% (239/295)
Glas, 200658Context: MammaPrint assay development through reanalysis of patients from the van't Veer21 and van de Vijver25 cohorts for which an RNA aliquot or the frozen specimen were available. A different reference RNA was used, as well as a different quantification method.The authors demonstrate for the first time that microarray technology can be used as a reliable diagnostic tool
Methods:
  • 162 lymph node negative, young breast cancer patients (<55 years), who did not received adjuvant therapy.
  • All 78 patients form the van't Veer 200221 cohort were available and used to reestablish the 70-genes expression profile, using the same procedures originally used and the same correlation to the signature (0.4).
  • 145 sample from the van de Vijver25 cohort (61 also in van't Veer21) were used and a higher correlation was used for classification purposes (0.55).
The MammaPrint assay performed similarly to the original 70-genes signature and is, therefore, an excellent tool to predict outcome of disease in breast cancer patients
Results:
  • Comparison between MammaPrint and the original 70-genes signature using all the 78 patients from the van't Veer21 series for the risk of developing a metastasis as first relapse event within 5 years from surgery:
 MammaPrint OR = 13.95 (95%CI = 3.9–44)
 Original 70-genes signature OR = 15, 95%CI = 2.1 to 19)
  • 7 patients out of 78 were classified differently by MammaPrint with respect to the original 70-genes signature:
 2 good prognosis patients were correctly classified by MammaPrint, but not by the original 70-genes signature
 1 poor prognosis patient was correctly classified by MammaPrint, but not by the original 70-genes signature
 2 poor prognosis patient s were misclassified by MammaPrint, but not by the original 70-genes signature
 2 good prognosis patients were misclassified by MammaPrint, but not by the original 70-genes signature
  • Comparison between MammaPrint and the original 70-genes signature using 145/151 LN negative patients from the van de Vijver25 series for the risk of developing a metastasis as first relapse event within 5 years from surgery:
 HR for MammaPrint = 5.6 (95%CI = 2.4–7.3, p value = 0.0001)
  • The probability of remaining free of metastases at 5 years was not different between MammaPrint and the original 70-genes signature, with a p value = 0.89 in log-rank test
  • Similar results were obtained for OS
van de Vijver, 200225Context: 70-genes signature partial validation using a gene expression array containing 25,000 features was used to evaluate the prognostic performance of the 70-genes signature in a consecutive series of 295 young women (< 53 years at diagnosis), with early stage (I–II), ER-positive or negative, node positive or negative breast cancer, who received or not adjuvant therapy (hormonal therapy, chemotherapy or both).The 70-genes signatures is a better predictor of the risk of distant metastases than standard clinical predictors
End point: distant metastases as first relapse event; patients were censored on the date of the last follow-up visit, death from causes other than breast cancer, the recurrence of local or regional disease, or the development of a second primary cancer, including contralateral breast cancer and Overall Survival (OS)
 In this study, 61 of the 78 patients of the van't Veer21 cohort were included.
Methods:
  • Kaplan-Meier analysis on the time form surgery to the first metastasis event or to censoring.
  • Survival curves were compared by the log-rank test.
  • Values are expressed as means ±SE, calculated according to the method of Tsiatis.
  • All SEs were calculated with use of the sandwich estimator.
  • Cox proportional-hazards regression analysis was used to investigate the association between the correlation coefficient (C1) and metastases by adjusting for other clinical variables (tumor grade, vascular invasion, and number of involved axillary's lymph-nodes: 0 versus 1 to 3 or 0 versus > 4).
  • The linearity of the relation between the relative hazard ratio and the diameter of the tumor, age, and ER level of expression was tested with use of the Wald test for nonlinear components of restricted cubic splines.
  • Hazard ratio proportionality was assessed by the method of Grambsch and Therneau.
  • Difference between the relative hazard ratio before and after five years of follow-up with respect to the 70-genes prognosis signature was done by the Wald test.
Results:
  • The 70-genes gene expression signature:
 Proved to be associated with age, tumor grade, ER status (P value<0.001), and tumor size (P value=0.012)
 Was not associated with lymph-nodes, vascular invasion, surgery, chemotherapy, hormonal therapy
  • OR for the various subgroups of patients in the consecutive series analyzed:
 78 node negative patients used in van't Veer 200221
  31 Poor signature with metastases
  18 Poor signature without metastases
  3 Good signature with metastases
  26 Good signature with out metastases
  OR = 15, (95%CI = 3.3–56, p value < 0.001)
 67 node-negative patients NOT used in van't Veer 2002:
  11 Poor signature with metastases
  23 Poor signature without metastases
  1 Good signature with metastases
  32 Good signature with out metastases
  OR = 15.3, (95%CI = 1.8–127, p value = 0.003)
 180 node-positive and negative patients NOT used in van't Veer 2002:
  39 Poor signature with metastases
  65 Poor signature without metastases
  3 Good signature with metastases
  73 Good signature with out metastases
  OR = 14.6, (95%CI = 4.3–50, p value < 0.0001)
  • Probability of metastases at 5yrs for all the patients:
 115 Good signature patients: 94.7±2.1SE
 180 Poor signature patients: 60.5±3.8SE
  • Probability of metastases at 10 yrs for all the patients
 115 Good signature patients: 85.2±4.3SE
 180 Poor signature patients: 50.6±4.5SE
  • Probability of metastases at 5 yrs for node negative patients:
 93 Good signature patients: 4±3.2SE
 91 Poor signature patients: 56.2±5.5SE
  • Probability of metastases at 10 yrs for node negative patients:
 60 Good signature patients: 86.8±4.8SE
 91 Poor signature patients: 44.1±6.3SE
  • Probability of metastases at 5 yrs for node positive patients
 55 Good signature patients: 95.2±2.6SE
 89 Poor signature patients: 66.3±5.2SE
  • Probability of metastases at 10 yrs for node positive patients
 55 Good signature patients: 82.7±7.8SE
 89 Poor signature patients: 56.7±6.4SE
  • OS at 5 yrs for all the patients:
 15 Good patients 97.4±1.5SE
 180 Poor patients 74.1±3.3SE
  • OS at 10 yrs for all the patients:
 115 Good patients 94.5±2.6SE
 180 Poor patients 54.6±4.4SE
  • OS at 5 yrs for node-negative patients:
 60 Good patients 96.7±2.3SE
 91 Poor patients 71.5±4.8SE
  • OS at 10 yrs for node-negative patients:
 60 Good patients 96.7±2.3SE
 91 Poor patients 49.6±6.1SE
  • OS at 5 yrs for node-positive patients:
 55 Good patients 98.2±1.8SE
 89 Poor patients 76.5±4.6SE
  • OS at 10 yrs for node-positive patients:
 55 Good patients 92.0±4.8SE
 89 Poor patients 59.5±6.3SE
  • HAZARD RATIO for metastases as first event over the entire follow-up period:
 All patients, HR = 5.1, (95%CI = 2.9–9.0, p value < 0.001)
 151 node negative patients, HR = 5.5, (95%CI = 2.5–12.2, p value < 0.001)
 144 mode positive patients, HR = 4.5, (95%CI = 2–10.2, p value < 0.001)
  • HAZARD RATIO for OS over the entire follow-up period:
 All patients = 8.6, (95%CI = 4–19, p value < 0.001)
  • In multivariate analysis the following factors resulted to be independent predictors of risk of metastases as first event within 5 years:
 70-genes signature, HR = 4.6, 95%CI = 2.3–9.2, p value < 0.001
 Tumor size, HR = 1.56, 95%CI = 1.22–2.0, p value < 0.001
 Chemotherapy, HR = 0.37, 95%CI = 0.2–0.66, p value < 0.001
  The other covariates used (age, grade, node status, vascular invasion, ER status, surgery, and hormonal therapy) were not associated with the risk of metastasis
  • Reclassification and comparison with St. Gallen and NIH criteria:
 The 70-genes signature classified more node negative patients in the Good prognosis group than both the St. Gallen and NIH criteria (40%, 15% and 7% respectively)
 43 St. Gallen High Risk group patients were re-classified in the Good signature group
 86 St. Gallen High Risk group patients were classified in the Poor signature group
 KM analysis showed that the 70-genes signature groups defined within the St. Gallen High Risk group are different, with a p value < 0.001
 17 St. Gallen Low Risk group patients were classified in the Good signature group
 5 St. Gallen Low Risk group patients were classified in the Poor signature group
 KM analysis showed that the 70-genes signature groups are PROBABLY different within the St. Gallen Low Risk group, with a p value < 0.11
 53 NIH High Risk patients were classified in the Good signature group
 87 NIH High Risk patients were classified in the Poor signature group
 KM analysis showed that the 70-genes signature groups are different within the NIH High Risk group, with a p value < 0.001
 7 NIH-Low Risk group patients were classified in the Good signature group
 4 NIH-Low Risk group patients were classified in the Poor signature group
 KM analysis showed that the 70-genes signature groups are different within NIH Low Risk group, with a p value < 0.05;
van't Veer, 200221Context: 70-genes gene expression signature development using a gene expression array containing 25,000 features was used to select genes associated with metastases free survival at 5 years from the surgery.The 70-genes signatures is a better predictor of the risk of distant metastases than standard clinical predictors;
End point: distant metastases as first relapse event; patients were censored on the date of the last follow-up visit, death from causes other than breast cancer, the recurrence of local or regional disease, or the development of a second primary cancer, including contralateral breast cancer.
Methods:
  • Kaplan-Meier analysis on the time form surgery to the first metastasis event or to censoring.
  • Survival curves were compared by the log-rank test.
  • Multivariate proportional-hazards regression analysis was used to investigate the association between the correlation coefficient and the time to metastases by adjusting for other clinical variables.
  • In the multivariate analysis the following predictors were used: tumor grade, ER status, PR status, tumor size, patient age, and angioinvasion.
  • Values were expressed as means ±SE, calculated according to the method of Tsiatis.
  • All SEs were calculated with use of the sandwich estimator.
Results:
  • Optimized 70-genes signature performance in the training set (78 patients):
 65/78 correct predictions (patients developed or not metastases within 5 years)
 13 misclassified patients:
  5 patients with poor prognosis were classified in the 70-gene Good prognosis group
  8 patients with good prognosis were classified in the 70-gene Poor prognosis group
  • Optimized 70-genes signature performance in the validation set (19 patients):
 17/19 correct predictions (patients developed or not metastases within 5 years)
 2 misclassified patients:
  1 patients with poor prognosis was classified in the 70-gene Good prognosis group
  1 patients with good prognosis in the 70-gene Poor prognosis group
  • ODDS RATIO to develop metastases within 5 years for a patient in the 70-genes Poor prognosis group:
 OR=28, (95%CI=7–107), P value = 0.00000001
 Leave-one-out cross-validated OR =15, (95%CI=4–56), P value=0.0000041
  • In univariate analysis the ‘leave-one-out’ cross-validated 70 genes signature proved to be a better predictor of recurrence than grade, size(>2cm), age (<40yrs), ER negative status, and angioinvasion
  • In multivariate analysis the 70-genes signature showed to be an independent predictor of metastases within 5 years (logistic regression OR = 18, (95%CI=3.3–94, P value = 0.00014)
  • Comparison of the 70 genes signatures with St. Gallen and NIH criteria for risk stratification:
 St. Gallen:
  64/78 patients were defined as at high risk of metastases (33/34 of the patients who did develop metastases; 31/44 who did not develop metastases)
 NIH:
  72/78 patients were defined as at high risk of metastases (32/34 of the patients who did develop metastases; 40/44 who did not develop metastases)
  70-genes signature:
 43/78 patients were defined as at high risk of metastases (31/34 of the patients who did develop metastases; 18/44 who did not develop metastases);

TRANSBIG= Breast International Group (BIG) network for improved treatment tailoring; OS= overall survival; RFS = relapse free survival; TTM = time to distant metastases; ROC= receiver operating characteristic; ER= estrogen receptor; NPI= Nottingham Prognostic Index; CI= confidence interval; RT-PCR= real time polymerase chain reaction; qRT-PCR= quantitative real time polymerase chain reaction; HR= hazard ratio; RNA= ribonucleic acid; OR= odds ratio; SE= standard error; NIH= National Institutes of Health; Kaplan Meier= Kaplan Meier; PR= partial relapse.

From: Appendix I: Evidence Tables

Cover of Impact of Gene Expression Profiling Tests on Breast Cancer Outcomes
Impact of Gene Expression Profiling Tests on Breast Cancer Outcomes.
Evidence Reports/Technology Assessments, No. 160.
Marchionni L, Wilson RF, Marinopoulos SS, et al.

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.