PD-L1 protein expression in breast cancer is rare, enriched in basal-like tumours and associated with infiltrating lymphocytes.

BACKGROUND: Expression of programmed death ligand 1 (PD-L1) in solid tumours has been shown to predict whether patients are likely to respond to anti-PD-L1 therapies. To estimate the therapeutic potential of PD-L1 inhibition in breast cancer, we evaluated the prevalence and significance of PD-L1 protein expression in a large collection of breast tumours. PATIENTS AND METHODS: Correlations between CD274 (PD-L1) copy number, transcript and protein levels were evaluated in tumours from 418 patients recruited to the METABRIC genomic study. Immunohistochemistry was used to detect PD-L1 protein in breast tumours in tissue microarrays from 5763 patients recruited to the SEARCH population-based study (N = 4079) and the NEAT randomised, controlled trial (N = 1684). RESULTS: PD-L1 protein data was available for 3916 of the possible 5763 tumours from the SEARCH and NEAT studies. PD-L1 expression by immune cells was observed in 6% (235/3916) of tumours and expression by tumour cells was observed in just 1.7% (66/3916). PD-L1 was most frequently expressed in basal-like tumours. This was observed both where tumours were subtyped by combined copy number and expression profiling [39% (17/44) of IntClust 10 i.e. basal-like tumours were PD-L1 immune cell positive; P < 0.001] and where a surrogate IHC-based classifier was used [19% (56/302) of basal-like tumours were PD-L1 immune cell positive; P < 0.001]. Moreover, CD274 (PD-L1) amplification was observed in five tumours of which four were IntClust 10. Expression of PD-L1 by either tumour cells or infiltrating immune cells was positively correlated with infiltration by both cytotoxic and regulatory T cells (P < 0.001). There was a nominally significant association between PD-L1 and improved disease-specific survival (hazard ratio 0.53, 95% confidence interval 0.26-1.07; P = 0.08) in ER-negative disease. CONCLUSIONS: Expression of PD-L1 is rare in breast cancer, markedly enriched in basal-like tumours and is correlated with infiltrating lymphocytes. PD-L1 inhibition may benefit the 19% of patients with basal-like tumours in which the protein is expressed. NEAT CLINICALTRIALSGOV: NCT00003577.

Association between CD8+ T-cell infiltration and breast cancer survival in 12,439 patients.

BACKGROUND: T-cell infiltration in estrogen receptor (ER)-negative breast tumours has been associated with longer survival. To investigate this association and the potential of tumour T-cell infiltration as a prognostic and predictive marker, we have conducted the largest study of T cells in breast cancer to date. PATIENTS AND METHODS: Four studies totalling 12 439 patients were used for this work. Cytotoxic (CD8+) and regulatory (forkhead box protein 3, FOXP3+) T cells were quantified using immunohistochemistry (IHC). IHC for CD8 was conducted using available material from all four studies (8978 samples) and for FOXP3 from three studies (5239 samples)-multiple imputation was used to resolve missing data from the remaining patients. Cox regression was used to test for associations with breast cancer-specific survival. RESULTS: In ER-negative tumours [triple-negative breast cancer and human epidermal growth factor receptor 2 (human epidermal growth factor receptor 2 (HER2) positive)], presence of CD8+ T cells within the tumour was associated with a 28% [95% confidence interval (CI) 16% to 38%] reduction in the hazard of breast cancer-specific mortality, and CD8+ T cells within the stroma with a 21% (95% CI 7% to 33%) reduction in hazard. In ER-positive HER2-positive tumours, CD8+ T cells within the tumour were associated with a 27% (95% CI 4% to 44%) reduction in hazard. In ER-negative disease, there was evidence for greater benefit from anthracyclines in the National Epirubicin Adjuvant Trial in patients with CD8+ tumours [hazard ratio (HR) = 0.54; 95% CI 0.37-0.79] versus CD8-negative tumours (HR = 0.87; 95% CI 0.55-1.38). The difference in effect between these subgroups was significant when limited to cases with complete data (P heterogeneity = 0.04) and approached significance in imputed data (P heterogeneity = 0.1). CONCLUSIONS: The presence of CD8+ T cells in breast cancer is associated with a significant reduction in the relative risk of death from disease in both the ER-negative [supplementary Figure S1, available at Annals of Oncology online] and the ER-positive HER2-positive subtypes. Tumour lymphocytic infiltration may improve risk stratification in breast cancer patients classified into these subtypes. NEAT ClinicalTrials.gov: NCT00003577.

Astronomical algorithms for automated analysis of tissue protein expression in breast cancer.

BACKGROUND: High-throughput evaluation of tissue biomarkers in oncology has been greatly accelerated by the widespread use of tissue microarrays (TMAs) and immunohistochemistry. Although TMAs have the potential to facilitate protein expression profiling on a scale to rival experiments of tumour transcriptomes, the bottleneck and imprecision of manually scoring TMAs has impeded progress. METHODS: We report image analysis algorithms adapted from astronomy for the precise automated analysis of IHC in all subcellular compartments. The power of this technique is demonstrated using over 2000 breast tumours and comparing quantitative automated scores against manual assessment by pathologists. RESULTS: All continuous automated scores showed good correlation with their corresponding ordinal manual scores. For oestrogen receptor (ER), the correlation was 0.82, P<0.0001, for BCL2 0.72, P<0.0001 and for HER2 0.62, P<0.0001. Automated scores showed excellent concordance with manual scores for the unsupervised assignment of cases to 'positive' or 'negative' categories with agreement rates of up to 96%. CONCLUSION: The adaptation of astronomical algorithms coupled with their application to large annotated study cohorts, constitutes a powerful tool for the realisation of the enormous potential of digital pathology.

Aurora kinase A outperforms Ki67 as a prognostic marker in ER-positive breast cancer.

BACKGROUND: Proliferation has emerged as a major prognostic factor in luminal breast cancer. The immunohistochemical (IHC) proliferation marker Ki67 has been most extensively investigated but has not gained widespread clinical acceptance. METHODS: We have conducted a head-to-head comparison of a panel of proliferation markers, including Ki67. Our aim was to establish the marker of the greatest prognostic utility. Tumour samples from 3093 women with breast cancer were constructed as tissue microarrays. We used IHC to detect expression of mini-chromosome maintenance protein 2, Ki67, aurora kinase A (AURKA), polo-like kinase 1, geminin and phospho-histone H3. We used a Cox proportional-hazards model to investigate the association with 10-year breast cancer-specific survival (BCSS). Missing values were resolved using multiple imputation. RESULTS: The prognostic significance of proliferation was limited to oestrogen receptor (ER)-positive breast cancer. Aurora kinase A emerged as the marker of the greatest prognostic significance in a multivariate model adjusted for the standard clinical and molecular covariates (hazard ratio 1.3; 95% confidence interval 1.1-1.5; P=0.005), outperforming all other markers including Ki67. CONCLUSION: Aurora kinase A outperforms other proliferation markers as an independent predictor of BCSS in ER-positive breast cancer. It has the potential for use in routine clinical practice.

PREDICT Plus: development and validation of a prognostic model for early breast cancer that includes HER2.

BACKGROUND: Predict (www.predict.nhs.uk) is an online, breast cancer prognostication and treatment benefit tool. The aim of this study was to incorporate the prognostic effect of HER2 status in a new version (Predict+), and to compare its performance with the original Predict and Adjuvant!. METHODS: The prognostic effect of HER2 status was based on an analysis of data from 10 179 breast cancer patients from 14 studies in the Breast Cancer Association Consortium. The hazard ratio estimates were incorporated into Predict. The validation study was based on 1653 patients with early-stage invasive breast cancer identified from the British Columbia Breast Cancer Outcomes Unit. Predicted overall survival (OS) and breast cancer-specific survival (BCSS) for Predict+, Predict and Adjuvant! were compared with observed outcomes. RESULTS: All three models performed well for both OS and BCSS. Both Predict models provided better BCSS estimates than Adjuvant!. In the subset of patients with HER2-positive tumours, Predict+ performed substantially better than the other two models for both OS and BCSS. CONCLUSION: Predict+ is the first clinical breast cancer prognostication tool that includes tumour HER2 status. Use of the model might lead to more accurate absolute treatment benefit predictions for individual patients.

Comparison of methods for handling missing data on immunohistochemical markers in survival analysis of breast cancer.

BACKGROUND: Tissue micro-arrays (TMAs) are increasingly used to generate data of the molecular phenotype of tumours in clinical epidemiology studies, such as studies of disease prognosis. However, TMA data are particularly prone to missingness. A variety of methods to deal with missing data are available. However, the validity of the various approaches is dependent on the structure of the missing data and there are few empirical studies dealing with missing data from molecular pathology. The purpose of this study was to investigate the results of four commonly used approaches to handling missing data from a large, multi-centre study of the molecular pathological determinants of prognosis in breast cancer. PATIENTS AND METHODS: We pooled data from over 11,000 cases of invasive breast cancer from five studies that collected information on seven prognostic indicators together with survival time data. We compared the results of a multi-variate Cox regression using four approaches to handling missing data - complete case analysis (CCA), mean substitution (MS) and multiple imputation without inclusion of the outcome (MI-) and multiple imputation with inclusion of the outcome (MI+). We also performed an analysis in which missing data were simulated under different assumptions and the results of the four methods were compared. RESULTS: Over half the cases had missing data on at least one of the seven variables and 11 percent had missing data on 4 or more. The multi-variate hazard ratio estimates based on multiple imputation models were very similar to those derived after using MS, with similar standard errors. Hazard ratio estimates based on the CCA were only slightly different, but the estimates were less precise as the standard errors were large. However, in data simulated to be missing completely at random (MCAR) or missing at random (MAR), estimates for MI+ were least biased and most accurate, whereas estimates for CCA were most biased and least accurate. CONCLUSION: In this study, empirical results from analyses using CCA, MS, MI- and MI+ were similar, although results from CCA were less precise. The results from simulations suggest that in general MI+ is likely to be the best. Given the ease of implementing MI in standard statistical software, the results of MI+ and CCA should be compared in any multi-variate analysis where missing data are a problem.

BCL2 in breast cancer: a favourable prognostic marker across molecular subtypes and independent of adjuvant therapy received.

BACKGROUND: Breast cancer is heterogeneous and the existing prognostic classifiers are limited in accuracy, leading to unnecessary treatment of numerous women. B-cell lymphoma 2 (BCL2), an antiapoptotic protein, has been proposed as a prognostic marker, but this effect is considered to relate to oestrogen receptor (ER) status. This study aimed to test the clinical validity of BCL2 as an independent prognostic marker. METHODS: Five studies of 11 212 women with early-stage breast cancer were analysed. Individual patient data included tumour size, grade, lymph node status, endocrine therapy, chemotherapy and mortality. BCL2, ER, progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) levels were determined in all tumours. A Cox model incorporating the time-dependent effects of each variable was used to explore the prognostic significance of BCL2. RESULTS: In univariate analysis, ER, PR and BCL2 positivity was associated with improved survival and HER2 positivity with inferior survival. For ER and PR this effect was time dependent, whereas for BCL2 and HER2 the effect persisted over time. In multivariate analysis, BCL2 positivity retained independent prognostic significance (hazard ratio (HR) 0.76, 95% confidence interval (CI) 0.66-0.88, P<0.001). BCL2 was a powerful prognostic marker in ER- (HR 0.63, 95% CI 0.54-0.74, P<0.001) and ER+ disease (HR 0.56, 95% CI 0.48-0.65, P<0.001), and in HER2- (HR 0.55, 95% CI 0.49-0.61, P<0.001) and HER2+ disease (HR 0.70, 95% CI 0.57-0.85, P<0.001), irrespective of the type of adjuvant therapy received. Addition of BCL2 to the Adjuvant! Online prognostic model, for a subset of cases with a 10-year follow-up, improved the survival prediction (P=0.0039). CONCLUSIONS: BCL2 is an independent indicator of favourable prognosis for all types of early-stage breast cancer. This study establishes the rationale for introduction of BCL2 immunohistochemistry to improve prognostic stratification. Further work is now needed to ascertain the exact way to apply BCL2 testing for risk stratification and to standardise BCL2 immunohistochemistry for this application.

Molecular characteristics of screen-detected vs symptomatic breast cancers and their impact on survival.

BACKGROUND: Several recent studies have shown that screen detection remains an independent prognostic factor after adjusting for disease stage at presentation. This study compares the molecular characteristics of screen-detected with symptomatic breast cancers to identify if differences in tumour biology may explain some of the survival benefit conferred by screen detection. METHODS: A total of 1379 women (aged 50-70 years) with invasive breast cancer from a large population-based case-control study were included in the analysis. Individual patient data included tumour size, grade, lymph node status, adjuvant therapy, mammographic screening status and mortality. Immunohistochemistry was performed on tumour samples using 11 primary antibodies to define five molecular subtypes. The effect of screen detection compared with symptomatic diagnosis on survival was estimated after adjustment for grade, nodal status, Nottingham Prognostic Index (NPI) and the molecular markers. RESULTS: Fifty-six per cent of the survival benefit associated with screen-detected breast cancer was accounted for by a shift in the NPI, a further 3-10% was explained by the biological variables and more than 30% of the effect remained unexplained. CONCLUSION: Currently known biomarkers remain limited in their ability to explain the heterogeneity of breast cancer fully. A more complete understanding of the biological profile of breast tumours will be necessary to assess the true impact of tumour biology on the improvement in survival seen with screen detection.

Weibull regression with Bayesian variable selection to identify prognostic tumour markers of breast cancer survival.

As data-rich medical datasets are becoming routinely collected, there is a growing demand for regression methodology that facilitates variable selection over a large number of predictors. Bayesian variable selection algorithms offer an attractive solution, whereby a sparsity inducing prior allows inclusion of sets of predictors simultaneously, leading to adjusted effect estimates and inference of which covariates are most important. We present a new implementation of Bayesian variable selection, based on a Reversible Jump MCMC algorithm, for survival analysis under the Weibull regression model. A realistic simulation study is presented comparing against an alternative LASSO-based variable selection strategy in datasets of up to 20,000 covariates. Across half the scenarios, our new method achieved identical sensitivity and specificity to the LASSO strategy, and a marginal improvement otherwise. Runtimes were comparable for both approaches, taking approximately a day for 20,000 covariates. Subsequently, we present a real data application in which 119 protein-based markers are explored for association with breast cancer survival in a case cohort of 2287 patients with oestrogen receptor-positive disease. Evidence was found for three independent prognostic tumour markers of survival, one of which is novel. Our new approach demonstrated the best specificity.