Time-dependent relevance of steroid receptors in breast cancer

Prognostic relation of body mass index on extended aromatase inhibition treatment in postmenopausal patients with estrogen receptor positive breast cancer: A retrospective analysis of the SOLE trial

BACKGROUND

Obesity is associated with a greater risk of developing distant recurrences in patients with estrogen receptor-positive (ER+) breast cancer. This association is however poorly investigated in patients treated with extended endocrine treatment (ET). We therefore evaluated the prognostic role of BMI in the SOLE trial, where postmenopausal patients, after having completed 4-6 years of adjuvant ET, were treated with 5 additional years of continuous or intermittent letrozole.

PATIENTS & METHODS

We considered the 3606 patients with ER+ /HER2- lymph node-positive BC with available BMI from the SOLE trial (NCT00553410). Distant-recurrence free interval (DRFI) was the main endpoint, and breast cancer-free interval (BCFI), disease-free survival (DFS) and overall survival (OS) secondary endpoints. Adjusted risk ratios (RR) for distant metastases were estimated with crude cumulative incidence models.

RESULTS

38.6 % of the patients were underweight or normal weight, 36.5 % overweight and 24.9 % obese. BMI was associated with age, tumor size, number of positive lymph nodes, menopausal status and type of prior ET. In the adjusted analyses, the prognostic value of BMI was dependent on prior ET and extended ET arm (second-order interaction p-value<0.001 for DRFI, BCFI and DFS, but not for OS). For instance, in patients treated with both a selective estrogen receptor modulator and an aromatase inhibitor in the first five years, obesity, as compared to normal-weight, was associated with better (RRDRFI=0.61, 95 %CI: 0.42-0.90) and worse (RRDRFI=2.31, 95 %CI: 1.41-3.78) outcomes in the adjusted models, in patients treated with continuous and intermittent letrozole in the extended ET, respectively.

CONCLUSION

We observed that the prognostic relation of BMI changes according to the type of adjuvant ET and mode of administration of extended AI. This warrants further investigation.

Case-Base Neural Network: Survival analysis with time-varying, higher-order interactions

In the context of survival analysis, data-driven neural network-based methods have been developed to model complex covariate effects. While these methods may provide better predictive performance than regression-based approaches, not all can model time-varying interactions and complex baseline hazards. To address this, we propose Case-Base Neural Networks (CBNNs) as a new approach that combines the case-base sampling framework with flexible neural network architectures. Using a novel sampling scheme and data augmentation to naturally account for censoring, we construct a feed-forward neural network that includes time as an input. CBNNs predict the probability of an event occurring at a given moment to estimate the full hazard function. We compare the performance of CBNNs to regression and neural network-based survival methods in a simulation and three case studies using two time-dependent metrics. First, we examine performance on a simulation involving a complex baseline hazard and time-varying interactions to assess all methods, with CBNN outperforming competitors. Then, we apply all methods to three real data applications, with CBNNs outperforming the competing models in two studies and showing similar performance in the third. Our results highlight the benefit of combining case-base sampling with deep learning to provide a simple and flexible framework for data-driven modeling of single event survival outcomes that estimates time-varying effects and a complex baseline hazard by design. An R package is available at https://github.com/Jesse-Islam/cbnn.

Time-varying prognostic effects of primary tumor sidedness and grade after curative liver resection for colorectal liver metastases

BACKGROUND

The veracity of the proportional hazards (PH) requirement is rarely scrutinized in most areas of cancer research, although fulfilment of this assumption underpins widely-used Cox survival models. We sought to critically appraise the existence of prognostic factors with time-dependent effects and to characterize their impact on survival among CLM patients.

METHODS

Consecutive patients who underwent liver resection with curative intent for CLM at the Singapore General Hospital were identified from a prospectively-maintained database. We evaluated PH of 55 candidate variables, and parameters which departed significantly from proportionality were included in Cox models that incorporated an interaction term to account for time-dependent effects. As sensitivity analyses, we fitted Weibull mixture 'cure' models to handle long plateaus in the tails of survival curves, and also analyzed the restricted mean survival time.

RESULTS

318 consecutive patients who underwent curative liver resection for CLM between Jan 2000 and Nov 2016 were included in this analysis. Hazard ratios for tumor grade (poorly-versus well- and moderately-differentiated) were found to decrease from 3.135 (95% CI: 1.637-6.003) at 12 months to 2.048 (95% CI: 1.038-4.042) after 24 months, and ceased to be significant at 26 months. Compared to left-sided tumors, a right-sided tumor location was found to portend worse prognosis for the first 10 months after resection but subsequently confer a survival benefit due to a crossing of survival curves. Corroborating this observation, long-term cure fractions were estimated to be 25.5% (95% CI: 17.4%-33.6%) and 34.2% (95% CI: 17.4%-50.9%) among patients with left-sided and right-sided primary disease respectively.

CONCLUSION

Primary tumor sidedness and grade appear to exert time-varying prognostic effects in CLM patients undergoing curative liver resection.

Clinical Characteristics and Outcomes of Single Versus Double Hormone Receptor-Positive Breast Cancer in 2 Large Databases

PURPOSE

To identify biologic and outcome differences between double hormone receptor (HR)-positive (dHR⁺, estrogen receptor (ER)⁺/progesterone receptor [PgR⁺]) and single HR-positive (sHR⁺, either ER⁺/PgR^- or ER^-/PgR⁺) breast cancer; and to explore whether hormone therapy (HT) response in HER2-negative breast cancer correlates with HR status.

PATIENTS AND METHODS

This retrospective study was conducted by using 2 large breast cancer databases: the Surveillance, Epidemiology, and End Results (SEER) database and the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) clinical data set. Cox regression analysis was used to estimate overall survival (OS) and breast cancer-specific survival (BCSS) among sHR⁺ and dHR⁺ patients.

RESULTS

In the SEER database, dHR⁺ patients had significantly longer OS and BCSS than ER⁺/PgR^- patients in short-term follow-up (OS: hazard ratio = 0.620; 95% confidence interval [CI], 0.590, 0.652; P < .001; BCSS: hazard ratio = 0.493; 95% CI, 0.462, 0.526; P < .001). Meanwhile, ER^-/PgR⁺ patients had younger age, larger tumor size, and higher disease grade than dHR⁺ and ER⁺/PgR^- patients. In patients who received HT, dHR⁺ patients had a more favorable OS than ER⁺/PgR^- patients (hazard ratio = 0.789; 95% CI, 0.635, 0.982; P = .034), and ER^-/PgR⁺ patients had a worse OS than ER⁺/PgR^- patients at 10 years' follow-up (hazard ratio = 7.991; 95% CI, 1.053, 60.644; P = .044). However, these groups had similar outcomes over longer periods.

CONCLUSION

In HER2-negative breast cancer, sHR⁺ patients are associated with relatively worse characteristics and worse short-term outcomes than dHR⁺ patients. Additionally, the outcome of patients receiving HT may differ according to the HR status. However, further studies are needed to confirm these conclusions.

Serum Markers and Prognosis in Locally Advanced Breast Cancer

Background

Locally advanced breast cancer (LABC) represents a heterogeneous subgroup of breast cancer with an often dismal outcome. Identifying prognostic factors has acquired great significance for the selection of optimal treatment in individual patients.

Methods

Between January 1993 and December 1997, 103 patients were treated in our institution with multimodality treatment consisting of neoadjuvant chemotherapy followed by surgery, adjuvant chemotherapy and radiotherapy; tamoxifen was added in hormone receptor-positive cases. In the search for prognostic factors well-established parameters (clinical, pathological and treatment-related) as well as new features with potential value (c-erbB-2, baseline serum levels of CA 15.3 and CEA) were included in the univariate and multivariate analysis.

Results

At a median follow-up of 92 months (range, 8-130), the estimated five-year cancer-specific overall survival (OS) and disease-free survival (DFS) were 71.34% and 57.7%, respectively. Among the 22 different variables studied, only 10 were significantly correlated with OS and DFS. In multivariate analysis five retained independent prognostic value for both OS and DFS: tumor grade, serum markers, features of inflammatory breast cancer (IBC), response to neoadjuvant chemotherapy and lymph node status. With cutoff values of 35 U/mL for CA 15.3 and 5 ng/mL for CEA, the probability of five-year OS (Cox hazard ratio 3.91, P = 0.0009) and DFS (Cox hazard ratio 2.40, P = 0.02) decreased from 78% to 52% and from 68% to 47%, respectively, when at least one of these markers was abnormal.

Conclusions

Baseline serum levels of CEA and CA 15.3 emerged from this study as strong independent predictors of outcome in LABC, whose value adds to other established prognostic factors such as postoperative nodal status, IBC, histological grade and response to neoadjuvant chemotherapy.

A Prediction Model for Breast Cancer Recurrence after Adjuvant Hormone Therapy

Hormone therapy with tamoxifen has long been the established adjuvant treatment for node-positive, estrogen–receptor-positive breast cancer in postmenopausal women. Since 30–40% of these patients fail to respond, reliable outcome prediction is necessary for successful treatment allocation. Using pathobiological variables (available in most clinical records: tumor size, nodal involvement, estrogen and progesterone receptor content) from 596 patients recruited at a comprehensive cancer center, we developed a prediction model which we validated in an independent cohort of 175 patients recruited at a general hospital. Calculated at 3 and 4 years of follow-up, the discrimination indices were 0.716 [confidence limits (CL) 0.641, 0.752] and 0.714 (CL 0.650, 0.750) for the training data, and 0.726 (CL 0.591, 0.769) and 0.677 (CL 0.580, 0.745) for the testing data. Waiting for more effective approaches from genomic and proteomic studies, a model based on consolidated pathobiological variables routinely assessed at relatively low costs may be considered as the reference for assessing the gain of new markers over traditional ones, thus substantially improving the conventional use of prognostic criteria.

Time-dependent and nonlinear effects of prognostic factors in nonmetastatic colorectal cancer

The survival risk following curative surgery for nonmetastatic colorectal cancer (CRC) may be over- or underestimated due to a lack of attention to nonlinear effects and violation of the proportional hazards assumption. In this paper, we aimed to detect and interpret the shape of time-dependent and nonlinear effects to improve the predictive performance of models of prognoses in nonmetastatic CRC patients. Data for nonmetastatic CRC patients diagnosed between 2004 and 2012 were obtained from the Surveillance Epidemiology End Results registry. Time-dependent and nonlinear effects were tested and plotted. A nonlinear model that used random survival forests was implemented. The estimated 5-year cancer-specific death rate was 17.95% (95% CI, 17.70-18.20%). Tumor invasion depth, lymph node status, age at diagnosis, tumor grade, histology and tumor site were significantly associated with cancer-specific death. Nonlinear and time-dependent effects on survival were detected. Positive lymph node number had a larger effect per unit of measurement at low values than at high values, whereas age at diagnosis showed the opposite pattern. Moreover, nonproportional hazards were detected for all covariates, indicating that the contributions of these risks to survival outcomes decreased over time. The nonlinear model predicted prognoses more accurately (C-index: 0.7934, 0.7933-0.7934) than did the Fine and Gray model (C-index: 0.7550, 0.7510-0.7583). The three-dimensional cumulative incidence curves derived from nonlinear model were used to identify the change points of the risk trends. It would be useful to implement these findings in treatment plans and follow-up surveillance in nonmetastatic CRC patients.

Comparative benefit from small tumour size and adjuvant chemotherapy: clues for explaining breast cancer mortality decline

Background

Breast cancer mortality steadily declined from the 1990s and this has been attributed to early detection and/or to improvements in therapy. Which of those two has had the greater impact is a subject of contention.

Methods

A database of 386 patients, enrolled in a randomized clinical trial on the effect of adjuvant chemotherapy (CMF), was analysed. The probabilities of recurrence and death were estimated by the Fine and Gray’s model and by the Cox model. Time dependent covariate and interaction effects were investigated by additive models. Absolute risk reductions (ARR) related to adjuvant treatment or to tumour size [diameter ≤ 2 cm (T1) or >2 cm (T2/T3)] were estimated.

Results

CMF-related reduction in recurrence emerges early, reaches a maximum level at 3 years and persists at a constant level thereafter. Tumour-size-related recurrence reduction, after a maximum at 3 years, displays a progressive regular reduction approaching zero. Patients with any tumour size, when given CMF, exhibit mortality reduction that displays an early regular increase and continues to a persistent plateau. In contrast, tumour-size-related mortality reduction reaches a maximum at 5–7 years and then regularly drops to very low values for patients of both trial arms.

Conclusions

Findings reveal that there is a different time-dependent benefit from chemotherapy and from smaller tumour size at diagnosis. The benefit from adjuvant chemotherapy is long-lasting for patients with any tumour size while the early benefit of diagnosing smaller tumours substantially decreases afterwards. Treatment improvements have probably had greater impact on the mortality reduction than mammography screening.

Cathepsin D as an indicator of clinical outcome in early breast carcinoma during the first 3 years of follow-up

AIM

The aim of this study was to evaluate clinical usefulness of cathepsin D status in early breast cancer during the first 3 years of follow-up.

PATIENTS & METHODS

The study included 226 patients with histologically verified, primary operable invasive early breast carcinomas. Concentrations of estrogen receptor (ER) and progesterone receptor (PR) in breast tumor cytosols were measured by use of the classical biochemical method. The concentration of three cathepsin D forms (52-, 48- and 34-kDa proteins) was determined by a radioimmunoassay

RESULTS

On the basis of differences in cathepsin D levels either within an ER(-)/PR(-) phenotype or between this and either ER(+)/PR(+) or ER(+)/PR(-) phenotypes, a concentration of 39 pmol/mg was determined as the cutoff value for distinguishing estrogen-regulated cathepsin D expression. Estrogen-regulated cathepsin D expression was recognized as a high-risk biomarker for low-risk (histological grade I) breast cancer patients and as a low-risk biomarker for high-risk patients (pN(+) pT2,3).

CONCLUSION

Determination of cathepsin D status in breast cancer might identify patients at different risk for relapse and might facilitate the selection of more or less aggressive adjuvant therapy for early breast cancer patients during the first 3 years of follow-up.

Trefoil factor 1 in early breast carcinoma: a potential indicator of clinical outcome during the first 3 years of follow-up

BACKGROUND

A role of an estrogen-regulated, autocrine motogenic factor was assumed to be a major biological role of trefoil factor 1 (TFF1) in breast cancer. TFF1 is regarded as a predictive factor for positive response to endocrine therapy in breast cancer patients. The aim of our study was to examine TFF1 level distribution in breast carcinomas in order to distinguish estrogen-independent from estrogen-dependent TFF1 expression and to evaluate clinical usefulness of TFF1 status in early breast cancer during the first 3 years of follow-up.

METHODS

The study included 226 patients with primary operable invasive early breast carcinomas for whom an equal, a 3-year follow-up was conducted. TFF1 levels as well as estrogen receptor (ER) and progesterone receptor (PR) levels were measured in cytosolic extracts of tumor samples by immunoradiometric assay or by use of classical biochemical method, respectively. Non-parametric statistical tests were applied for data analyses.

RESULTS

Statistical analysis revealed that TFF1 levels were significantly higher in premenopausal patients (p=0.02), or in tumors with: lower histological grade (p<0.001), positive ER or PR status (p<0.001, in both cases). On the basis of TFF1 level distribution between ER-negative and ER-positive postmenopausal patients with tumors of different histological grade, 14 ng/mg was set as the cut-off value to distinguish estrogen-independent from estrogen-dependent TFF1 expression in breast cancer. Depending on menopausal and PR status, positive TFF1 status identified patients at opposite risk for relapse among ER-positive patients with grade II tumors. Among ER- and PR-positive premenopausal patients with grade II tumors, TFF1 status alone identified patients at opposite risk for relapse.

CONCLUSIONS

Determination of TFF1 status might identify patients at different risk for relapse and help in making decision on administering adjuvant therapy for early breast cancer patients during the first 3 years of follow-up.

Reporting recommendations for tumor marker prognostic studies (REMARK): explanation and elaboration

BACKGROUND

The Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) checklist consists of 20 items to report for published tumor marker prognostic studies. It was developed to address widespread deficiencies in the reporting of such studies. In this paper we expand on the REMARK checklist to enhance its use and effectiveness through better understanding of the intent of each item and why the information is important to report.

METHODS

REMARK recommends including a transparent and full description of research goals and hypotheses, subject selection, specimen and assay considerations, marker measurement methods, statistical design and analysis, and study results. Each checklist item is explained and accompanied by published examples of good reporting, and relevant empirical evidence of the quality of reporting. We give prominence to discussion of the 'REMARK profile', a suggested tabular format for summarizing key study details.

SUMMARY

The paper provides a comprehensive overview to educate on good reporting and provide a valuable reference for the many issues to consider when designing, conducting, and analyzing tumor marker studies and prognostic studies in medicine in general. To encourage dissemination of the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK): Explanation and Elaboration, this article has also been published in PLoS Medicine.

Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK): explanation and elaboration

The REMARK (Reporting Recommendations for Tumor Marker Prognostic Studies) guideline includes a checklist which aims to improve the reporting of these types of studies. Here, we expand on the REMARK checklist to enhance its use and effectiveness through better understanding of the intent of each item and why the information is important to report. Each checklist item of the REMARK guideline is explained in detail and accompanied by published examples of good reporting. The paper provides a comprehensive overview to educate on good reporting and provide a valuable reference of issues to consider when designing, conducting, and analyzing tumor marker studies and prognostic studies in medicine in general.

Histology-specific nomogram for primary retroperitoneal soft tissue sarcoma

BACKGROUND

This study was conducted to develop a histology-specific nomogram to predict postoperative overall survival (OS) at 5 and 10 years in primary retroperitoneal soft tissue sarcoma (STS).

METHODS

Data registered at a single institution (National Cancer Institute, Milan, Italy) prospective sarcoma database were used. In the present analysis, patients with primary localized retroperitoneal STS resected with curative intent between 1985 and 2007 were included. A parametric piecewise exponential survival multivariate model was used for nomogram development, and internal validation was performed with standard methodologies. Known prognostic variables, such as age, tumor burden, histologic variant (as reviewed by a sarcoma pathologist), grade, and surgical margins were considered as putative predictors.

RESULTS

Among the 192 patients analyzed, within 10 years from surgery, 114 patients were alive, with a median follow-up time of 55 months (interquartile range, 25-104 months). Among the investigated factors, only histologic subtype did not reach significance at the 10% level. The relative hazard increased while increasing tumor size up to about 25 cm, and decreased thereafter.

CONCLUSIONS

A histology-specific nomogram for retroperitoneal STS is now available. It can be used for better assessing the risk of the single patient and then making individualized decisions within the specific subset of retroperitoneal sarcomas. Cross-check external validation should be performed.

Recurrence and mortality dynamics for breast cancer patients undergoing mastectomy according to estrogen receptor status: different mortality but similar recurrence

(Cancer Sci 2010; 101: 826-830) The purpose was to ascertain whether the recurrence risk patterns for patients with estrogen receptor (ER)-positive (P) and ER-negative (N) breast cancer support the ER-related clinical divergence suggested by the observed different mortality patterns and gene expression profiles. Both recurrence and death were considered in a series of 771 patients undergoing mastectomy. ER status was available for 539 patients. The hazard rates for recurrence and mortality throughout 15 years of follow-up were assessed. The recurrence dynamics displays a bimodal pattern for both ERP and ERN tumors with comparable peak timings. The two curves cross during the 3rd year. By contrast, the mortality dynamics are definitely different for ERP and ERN tumors: during the early follow-up period ERN patients have their highest mortality risk, while ERP patients have their lowest mortality risk. The two curves cross during the 5th year. In spite of the different mortality dynamics, the recurrence dynamics do not demonstrate a major distinction in timing between ERP and ERN breast cancers, suggesting that the metastasis development process following mastectomy is apparently similar for both ER categories. The observed differences in the mortality risk are plausibly attributable to ER-related factors influencing the clinical course from recurrence to death. These clinical findings apparently contradict the occurrence of two different types of breast cancer, notwithstanding the distinct epidemiological, clinical, and molecular features linked to ERP and ERN tumors, although ER levels may concur to establish the event risk levels.

time dependency of prognostic factors in patients with stage II osteosarcomas

Although several clinicopathologic parameters may be related to metastasis-free survival in osteosarcoma, the importance of prognostic factors with time is largely unknown. Therefore, we asked whether the clinical characteristics of patients with osteosarcoma with metastasis after 2 years differed from those of patients with an earlier metastatic event. We retrospectively reviewed 420 patients with Stage IIB osteosarcoma treated with surgery and chemotherapy. The minimum followup was 3 months (mean, 71.4 months; range, 3-257 months). Among the 420 patients, 167 patients had a metastatic event. A large proportion (43%) of the 35 good responders showed metastasis after 2 years, whereas 24% of the 96 poor responders showed late metastasis. Histologic response and chondroblastic subtype showed time-related changes. Poor histologic response showed a gradual decline in prognostic value and lost its importance after 2 years. Most metastatic events occurred within 2 years after diagnosis; therefore, we recommend new predictive strategies and followup protocol for patients experiencing late relapse.

A new proposal for multivariable modelling of time-varying effects in survival data based on fractional polynomial time-transformation

The Cox proportional hazards model has become the standard for the analysis of survival time data in cancer and other chronic diseases. In most studies, proportional hazards (PH) are assumed for covariate effects. With long-term follow-up, the PH assumption may be violated, leading to poor model fit. To accommodate non-PH effects, we introduce a new procedure, MFPT, an extension of the multivariable fractional polynomial (MFP) approach, to do the following: (1) select influential variables; (2) determine a sensible dose-response function for continuous variables; (3) investigate time-varying effects; (4) model such time-varying effects on a continuous scale. Assuming PH initially, we start with a detailed model-building step, including a search for possible non-linear functions for continuous covariates. Sometimes a variable with a strong short-term effect may appear weak or non-influential if 'averaged' over time under the PH assumption. To protect against omitting such variables, we repeat the analysis over a restricted time-interval. Any additional prognostic variables identified by this second analysis are added to create our final time-fixed multivariable model. Using a forward-selection algorithm we search for possible improvements in fit by adding time-varying covariates. The first part to create a final time-fixed model does not require the use of MFP. A model may be given from 'outside' or a different strategy may be preferred for this part. This broadens the scope of the time-varying part. To motivate and illustrate the methodology, we create prognostic models from a large database of patients with primary breast cancer. Non-linear time-fixed effects are found for progesterone receptor status and number of positive lymph nodes. Highly statistically significant time-varying effects are present for progesterone receptor status and tumour size.

Ki-67, p53, ER receptors, ploidy and S phase as long-term prognostic factors in T1 node-negative breast cancer

BACKGROUND

The aim of the present study was to evaluate a series of biomarkers with regard to long-term prognostic value in patients with T1 (< or =2 cm) node-negative breast cancer.

METHOD

The prognostic value of Ki-67, p53, oestrogen receptor (ER) immunohistochemical labelling, flow-cytometric S phase fraction and ploidy was evaluated in 212 patients with pT1N0M0 breast cancer. The median follow-up time was 15.9 years (range 0.2-27.2 years).

RESULTS

In an analysis of breast cancer-specific survival up to 5 years, high Ki-67 (> or =10%; p = 0.002), high p53 (> or =20%; p = 0.01), negative ER (<30%; p = 0.01) as well as aneuploidy of the tumour (p = 0.02) were significant prognostic factors. When the follow-up was extended to 10 years, only Ki-67 (p = 0.03) was significantly associated with outcome and beyond 15 years none of the studied markers provided significant prognostic information when analyzed separately. There was a weak but significant difference in long-term survival when patients with a combination of high Ki-67 (> or =10%), high SPF (>3%) and high p53 (> or =20%) were compared to patients with other combinations (p = 0.03).

CONCLUSION

According to the results of our series, it seems that several prognostic markers which are associated with short-term survival (< or =5 years) in pT1N0M0 breast cancer may not be significant predictors of long-term (>15 years) breast cancer-specific survival.

Patient-related predictors of implant failure after primary total hip replacement in the initial, short- and long-terms. A nationwide Danish follow-up study including 36,984 patients

We examined the association between patient-related factors and the risk of initial, short- and long-term implant failure after primary total hip replacement. We used data from the Danish Hip Arthroplasty Registry between 1 January 1995 and 31 December 2002, which gave us a total of 36 984 patients. Separate analyses were carried out for three follow-up periods: 0 to 30 days, 31 days to six months (short term), and six months to 8.6 years after primary total hip replacement (long term). The outcome measure was defined as time to failure, which included re-operation with open surgery for any reason. Male gender and a high Charlson co-morbidity index score were strongly predictive for failure, irrespective of the period of follow-up. Age and diagnosis at primary total hip replacement were identified as time-dependent predictive factors of failure. During the first 30 days after primary total hip replacement, an age of 80 years or more and hip replacement undertaken as a sequela of trauma, for avascular necrosis or paediatric conditions, were associated with an increased risk of failure. However, during six months to 8.6 years after surgery, being less than 60 years old was associated with an increased risk of failure, whereas none of the diagnoses for primary total hip replacement appeared to be independent predictors.

Artificial neural network for the joint modelling of discrete cause-specific hazards

OBJECTIVE

Artificial neural network (ANN) based regression methods have been introduced for modelling censored survival data to account for complex prognostic patterns. In the framework of ANN extensions of generalized linear models for survival data, PLANN is a partial logistic ANN, suitable for smoothed discrete hazard estimation as a function of time and covariates. An extension of PLANN for competing risks analysis (PLANNCR) is now proposed for discrete or grouped survival times, resorting to the multinomial likelihood.

METHODS AND MATERIALS

PLANNCR is built by assigning input nodes to the explanatory variables with the time interval treated as an ordinal variable. The logistic function is used as activation for the hidden nodes of the network, whereas the softmax, which corresponds to the canonical link of generalized linear models for polytomous regression, is adopted for multiple output nodes, to provide a smoothed estimation of discrete conditional event probabilities for each event. The Kullback-Leibler distance is used as error function for the target vectors, amounting to half of the deviance of a multinomial logistic regression model. PLANNCR can jointly model non-linear, non-proportional and non-additive effects on cause-specific hazards (CSHs). The degree of smoothing is modulated by the number of hidden nodes and penalization of the error function (weight decay). Model optimisation is achieved by quasi-Newton algorithms, while non-linear cross-validation (NCV) and the Network Information Criterion (NIC) were adopted for model selection. PLANNCR was applied to data on 1793 women with primary invasive breast cancer, histologically N-, who underwent surgery at the Milan Cancer Institute between 1981 and 1986.

RESULTS

Differential effects of covariates and time on the shape of the CSH for the three main failure causes, namely intra-breast tumor recurrences, distant metastases and contralateral breast cancer, have been enlightened.

CONCLUSIONS

PLANNCR can be suitably adopted in an exploratory framework for a thorough evaluation of the disease dynamics in the presence of competing risks.

Molecular Subtyping of Breast Cancer from Traditional Tumor Marker Profiles Using Parallel Clustering Methods

PURPOSE

Recent small-sized genomic studies on the identification of breast cancer bioprofiles have led to profoundly dishomogenous results. Thus, we sought to identify distinct tumor profiles with possible clinical relevance based on clusters of immunohistochemical molecular markers measured on a large, single institution, case series.

EXPERIMENTAL DESIGN

Tumor biological profiles were explored on 633 archival tissue samples analyzed by immunohistochemistry. Five validated markers were considered, i.e., estrogen receptors (ER), progesterone receptors (PR), Ki-67/MIB1 as a proliferation marker, HER2/NEU, and p53 in their original scale of measurement. The results obtained were analyzed by three different clustering algorithms. Four different indices were then used to select the different profiles (number of clusters).

RESULTS

The best classification was obtained creating four clusters. Notably, three clusters were identified according to low, intermediate, and high ER/PR levels. A further subdivision in two biologically distinct subtypes was determined by the presence/absence of HER2/NEU and of p53. As expected, the cluster with high ER/PR levels was characterized by a much better prognosis and response to hormone therapy compared to that with the lowest ER/PR values. Notably, the cluster characterized by high HER2/NEU levels showed intermediate prognosis, but a rather poor response to hormone therapy.

CONCLUSIONS

Our results show the possibility of profiling breast cancers by means of traditional markers, and have novel clinical implications on the definition of the prognosis of cancer patients. These findings support the existence of a tumor subtype that responds poorly to hormone therapy, characterized by HER2/NEU overexpression.

The importance of temporal effects in evaluating the prognostic impact of joint ERPR expression in premenopausal women with node-positive breast cancer

Although there is abundant information about the independent effects of estrogen receptor (ER) and progesterone receptor (PR) on outcomes of breast cancer, comparatively little is known about the impact of joint (ER+PR+, ER+PR-, ER-PR+ and ER-PR-) ERPR expression. The purpose of this study was to evaluate the prognostic relevance of joint ERPR expression to progression free survival (PFS). Data from 710 patients with a median follow-up of 119 months has been analyzed retrospectively. Our results indicate that the effect of the ER+PR+ phenotype on PFS was significantly time-dependent (p<0.0001); favorable in the first 3 years of follow-up (HR=0.67, p=0.0175) compared to ER-PR- phenotype, but unfavorable during the later follow-up period (HR=2.89, p=0.0006). Similar patterns were also observed for ER+PR- and ER-PR+ phenotypes, but the effect did not reach statistical significance. In the tree-based analysis, we found that, among patients with more than 4 positive nodes and age greater than 40, those with ER-PR+ tumors had the worst PFS ( p=0.025), and among patients with 1--3 positive nodes and stage of T1 and T2, those with ER+PR- had the worst outcome ( p=0.006). Our results demonstrate that failure to recognize the time-varying effect of the steroid hormonal receptors can obscure their role in the prognosis of breast cancer. We also provide more evidence to support the concept that ER-PR+ is a real group representing a distinct clinical entity.

Multiple correspondence analysis in S-PLUS

Multiple correspondence analysis (MCA) is a multivariate method for analyzing multidimensional contingency tables. General software procedures to perform MCA are available. Among them SAS Proc CORRESP, SPAD CORMU procedure and the mca function of the MASS library in S-PLUS are probably the most used. However, CORRESP and CORMU output is different from that of mca function. The aim of this short note is showing how to obtain from mca function results compatible with those achieved with SAS or SPAD. A modified code is proposed in order to obtain the same coordinate system computed by SAS and SPAD. Moreover, the computation of the contributions of the levels of the factors to the inertia explained by each axis, the squared cosine of each factor level and the re-evaluation of the inertia explained by each axis have been added in order to improve the interpretations of the results of the decomposition.

[Molecular alterations in breast cancer: clinical implications and new analytical tools]

PURPOSE

Breast cancer is a complex and heterogeneous disease resulting from various molecular alterations, the identification of which should have profound impact on the management of patients.

CURRENT KNOWLEDGE AND KEY POINTS

The discovery of germline mutations within breast cancer susceptibility genes, such as BRCA1 and BRCA2, which are associated with a major risk of breast cancer during lifetime, has improved the assessment of the individual risk toward the disease, allowing appropriate strategies of screening and prevention. The identification of key molecular actors in the mammary oncogenesis may help to better assess the prognosis of the disease, while providing new therapeutic targets. Large-scale molecular technologies, which allow simultaneous assessment of a high number of molecular parameters in a single assay, should provide new tools to tackle complexity and heterogeneity of breast cancer. Hence, by examining transcriptional profiles of breast cancer using DNA microarrays, it was possible to reveal new prognostic tumor subgroups, previously indistinguishable. Further improvements are awaited with the recent development of high throughput and large-scale technologies investigating the tumor proteome.

PROSPECTS AND PROJECTS

Precise knowledge of molecular alterations involved in each individual breast cancer will allow more effective and less toxic, tailored therapies.

Quantitative target sizes for breast tumor detection prior to metastasis: a prerequisite to rational design of 4D scanners for breast screening

It is important to determine a breast cancer tumor target size for new screening equipment and molecular detection. Records of women aged 40-69 years diagnosed in 1988-1997 with a nonmetastasized, node-negative, or node-positive T1-stage breast cancer were abstracted from the Surveillance, Epidemiology, and End Results (SEER) public-use database. The linear, Gompertzian, lognormal, and power-exponential models of the effect of tumor size on breast cancer specific mortality were compared using corresponding transforms of size in multivariate Cox proportional hazard models. Criteria for comparison were the linearization of the size transforms and the Nagelkerke R2N index for the Cox models. Our results show that the assumption of a linear effect of tumor size was rejected by the linearity test (P=0.05). The Gompertzian, lognormal, and power-exponential transforms satisfied the test with P-values of 0.08, 0.29, and 0.14, respectively. The corresponding R2N were 0.08410, 0.08420, and 0.08414, respectively, showing a marginally best fit with the lognormal model, which was selected as a model for small tumors. The lognormal function with unadjusted crude death rates gave a lognormal-location parameter of 25 and shape parameter of 1.7, while the corresponding values in multivariate models were 18 and 2, respectively. The derivation of the lognormal model indicates tumor growth acceleration starting at 3 mm (unadjusted crude data) or 2 mm (multivariate model). The breast cancer tumor target size for screening equipment, whether by imaging or molecular detection, is therefore 2 mm.

Prognostic factors affecting the natural history of node-negative breast cancer

PURPOSE

We undertook a natural history investigation of a broad selection of prognostic factors in a cohort of women with node-negative breast cancer.

PATIENTS AND METHODS

The cohort consisted of 415 consecutive histologic node-negative (T1-3, M0) patients, operated on for primary breast cancer at Women's College Hospital, Toronto, Canada, between 1977 and 1986. Only 7% of these patients were given adjuvant systemic therapy; further, for the 48% of women who underwent lumpectomy, only 29% received adjuvant radiotherapy to the breast. Paraffin-embedded tumour tissue was available for the majority of patients. The following factors were examined for their univariate and multivariate effects on time to recurrence outside the breast (DFI) and survival from breast cancer (DSS): age, weight, tumour size, estrogen receptor, progesterone receptor, histologic type, tumour grade, nuclear grade, lymphovascular invasion, overexpression of neu oncoprotein, DNA ploidy, % cells in S-phase, and adjuvant therapy. Multivariate analyses utilized a Cox model with a step-wise factor selection for the 260 patients with complete information.

RESULTS

A worse prognosis was indicated when there was lymphovascular invasion (for DFI, p < 0.001; for DSS, p = 0.0046), high %S-phase (for DFI, p = 0.08; for DSS, p = 0.02), high tumour grade (for DFI, p = 0.02; for DSS, p = 0.03), and overexpression of neu oncoprotein (for DSS, p = 0.07).

CONCLUSIONS

In our natural history investigation, two factors, lymphovascular invasion and tumour grade, are of particular interest since they may be readily incorporated into clinical practice. Overexpression of neu oncoprotein may also play a role in determining prognosis for women administered adjuvant systemic therapy.

Modeling the effect of tumor size in early breast cancer

SUMMARY BACKGROUND DATA

The purpose of this study was to determine the type of relationship between tumor size and mortality in early breast carcinoma.

METHODS

The data was abstracted from 83,686 cases registered in the Surveillance, Epidemiology, and End Results Program of women diagnosed with primary breast carcinoma between 1988 and 1997 presenting with a T1-T2 lesion and no metastasis in whom axillary node dissection was performed: 58,070 women were node-negative (N0) and 25,616 were node-positive (N+). End point was death from any cause. Tumor size was modeled as a continuous variable by proportional hazards using a generalized additive models procedure.

RESULTS

Functionally, a Gompertzian expression exp(-exp(-(size-15)/10)) provided a good fit to the effect of tumor size (in millimeters) on mortality, irrespective of nodal status. Quantitatively, for tumor size between 3 and 50 mm, the increase of crude cumulative death rate (number of observed deaths divided by the number of patients at risk) increased with size from 10% to 25% for N0 and from 20% to 40% for N+.

CONCLUSIONS

The functional relationship of tumor size with mortality is concordant with current knowledge of tumor growth. However, its qualitative and quantitative independence of nodal status is in contradiction with the prevailing concept of sequential disease progression from primary tumor to regional nodes. This argues against the perception that nodal metastases are caused by the primary tumor.

Circulating biomarkers from tumour bulk to tumour machinery: promises and pitfalls

In this paper, we provide a working classification for circulating biomarkers according to their potential clinical application. We broadly divided biomarkers into four groups: (i) biomarkers of cancer risk, (ii) biomarkers of tumour-host interactions, (iii) biomarker of tumour burden, and (iv) function-related biomarkers. We hope this classification will provide a framework to which the results of future studies can be added. We also discuss the promises and pitfalls in the optional use of biomarkers in oncology.

A time-dependent discrimination index for survival data

To derive models suitable for outcome prediction, a crucial aspect is the availability of appropriate measures of predictive accuracy, which have to be usable for a general class of models. The Harrell's C discrimination index is an extension of the area under the ROC curve to the case of censored survival data, which owns a straightforward interpretability. For a model including covariates with time-dependent effects and/or time-dependent covariates, the original definition of C would require the prediction of individual failure times, which is not generally addressed in most clinical applications. Here we propose a time-dependent discrimination index Ctd where the whole predicted survival function is utilized as outcome prediction, and the ability to discriminate among subjects having different outcome is summarized over time. Ctd is based on a novel definition of concordance: a subject who developed the event should have a less predicted probability of surviving beyond his/her survival time than any subject who survived longer. The predicted survival function of a subject who developed the event is compared to: (1) that of subjects who developed the event before his/her survival time, and (2) that of subjects who developed the event, or were censored, after his/her survival time. Subjects who were censored are involved in comparisons with subjects who developed the event before their observed times. The index reduces to the previous C in the presence of separation between survival curves on the whole follow-up. A confidence interval for Ctd is derived using the jackknife method on correlated one-sample U-statistics.The proposed index is used to evaluate the discrimination ability of a model, including covariates having time-dependent effects, concerning time to relapse in breast cancer patients treated with adjuvant tamoxifen. The model was obtained from 596 patients entered prospectively at Istituto Nazionale per lo Studio e la Cura dei Tumori di Milano (INT). The model discrimination ability was validated on an independent testing data set of 175 patients provided by Centro Regionale Indicatori Biochimici di Tumore (CRIBT) in Venice.

Time dependence of prognostic factors for patients with soft tissue sarcoma: a Scandinavian Sarcoma Group Study of 338 malignant fibrous histiocytomas

BACKGROUND

Prognostic factors for metastasis in soft tissue sarcoma govern decisions regarding adjuvant treatment. However, the significance of initial tumor-related prognostic factors over time is largely unknown.

METHODS

The current study included 338 patients with malignant fibrous histiocytoma (MFH) of the extremities or the trunk wall whose tumors were reviewed by the Scandinavian Sarcoma Pathology Review Group. Of these 338 patients, 329 (97%) had high-grade tumors. The median follow-up period was 7 years. Metastases occurred in 110 of 338 of patients after a median follow-up period of 14 months, with roughly one-third (32 of 110) occurring after 2 years. The authors investigated the prognostic significance of tumor size, tumor depth, histologic grade, microscopic tumor necrosis, vascular invasion, mitotic rate, and local tumor recurrence at various time intervals using metastases as an endpoint.

RESULTS

On univariate analysis, all investigated factors were found to be correlated with metastases for the entire follow-up period and also for the first 2 years of follow-up; beyond this time point, only size, tumor depth, and local recurrence were significant. On multivariate analysis, necrosis and local tumor recurrence were significant for the entire follow-up duration and also for the first 2 years of follow-up, whereas only tumor depth and local recurrence were significant beyond 2 years of follow-up. For all initial factors, the annual metastasis risks in the high-risk and low-risk groups converged to < 0.1 after 2 years and to near 0 after 5 years.

CONCLUSIONS

Prognostic factors for metastasis in MFH were time dependent. The predictive value of the initial prognostic factors was limited to the first 2 years of follow-up. The lack of observed prognostic value beyond 2 years of follow-up probably was attributable to heterogeneity within risk categories as a result of measurement errors and unknown biologic variations.

Prognosis in node-negative primary breast cancer: a neural network analysis of risk profiles using routinely assessed factors

BACKGROUND

The present study investigated complex time-dependent effects of routinely assessed factors on the risk of breast cancer recurrence over follow-up time, with a partial logistic artificial neural network (PLANN) model.

PATIENTS AND METHODS

PLANN was applied to data from 1793 patients with node-negative breast cancer, not submitted to any adjuvant treatment and with a minimal potential follow-up of 10 years.

RESULTS

The shape of the hazard function changed according to histology, which showed a time-dependent effect, partly modulated by estrogen receptors (ERs). Age and progesterone receptors (PgR) showed protective effects; the latter was more evident for short follow-up and high ER values. Tumour size and ER content showed time-dependent unfavourable effects at early and long follow-up times, respectively. Predicted values of disease recurrence probability at 2 years of follow-up showed that low steroid-receptor content, young age and large tumour size were associated with the highest risk of relapse. Although the oldest patients with high ER content seem to be those most protected overall, high risk predictions tend to spread also to higher steroid-receptor contents, intermediate ages and small tumour size, with an increase in follow-up time.

CONCLUSION

PLANN with suitable visualisation techniques provided thorough insights into the dynamics of breast cancer recurrence for improving individual risk staging of node-negative breast cancer patients.

Vascular endothelial growth factor in node-positive breast cancer patients treated with adjuvant tamoxifen

In 212 postmenopausal women with node-positive oestrogen receptor-positive (ER(LBA)) breast cancer subjected to radical surgery and adjuvant tamoxifen, the risk of 6-year relapse increased with increasing values of intratumoral vascular endothelial growth factor (VEGF) in patients whose tumours had a low/intermediate ER(LBA) content compared to patients with high-ER(LBA) tumours. These findings indicate that tumour progression, activated or sustained by high VEGF levels, may be counteracted in high-ER(LBA) cancers by tamoxifen, which in contrast fails to contrast the metastatic potential in low-ER(LBA) tumours.

Progesterone receptor status significantly improves outcome prediction over estrogen receptor status alone for adjuvant endocrine therapy in two large breast cancer databases

PURPOSE

To determine whether progesterone receptor (PgR) status provides additional value to estrogen receptor (ER) status and improves prediction of benefit from endocrine treatment among patients with primary breast cancer.

PATIENTS AND METHODS

Clinical outcomes of patients in two large databases were analyzed as a function of steroid receptor status. The first database (PP), contained 3,739 patients who did not receive any systemic adjuvant therapy and 1,688 patients who received adjuvant endocrine therapy but no chemotherapy. The second database (SPORE), contained 10,444 patients who received adjuvant endocrine therapy but no chemotherapy. Biochemical ER and PgR assays were identically performed in two different central laboratories.

RESULTS

In univariate and multivariate analyses, the prognostic significance of PgR status among systemically untreated patients is modest. Among endocrine-treated patients, however, multivariate analyses, including lymph-node involvement, tumor size, and age, demonstrate that PgR status is independently associated with disease-free and overall survival. For recurrence, the reduction in relative risk (RR) was 25% for ER-positive/PgR-negative patients and 53% for ER-positive/PgR-positive patients, compared with ER-negative/PgR-negative patients (P <.0001, PP patients). Patients with ER-positive/PgR-negative tumors have a reduction in RR of death of 30% (SPORE patients) and 38% (PP patients), compared with patients with ER-negative/PgR-negative tumors (P <.0001). For ER-positive/PgR-positive tumors, the reduction of the risk of death was greater than 46% in SPORE patients and 58% in PP patients, indicating that ER-positive/PgR-positive patients derive more benefit from endocrine therapy (P <.0001).

CONCLUSION

When accurately measured, PgR status is an independent predictive factor for benefit from adjuvant endocrine therapy. Therefore, PgR status should be taken into account when discussing RR reductions expected from endocrine treatment with individual patients.

Estrogen receptors and distinct patterns of breast cancer relapse

BACKGROUND

We conducted an analysis of prospectively collected data to compare the clinical behavior of ER-negative versus ER-positive tumors with respect to rates and sites of recurrence.

METHODS

A total of 647 patients with operable stage II or III breast cancer were enrolled in two consecutive adjuvant therapy protocols conducted between 1980 and 1986. The correlations between ER status and time to first recurrence, site of first recurrence, and time to breast cancer death were assessed on 558 (86%) patients with available ER status data using hazard function and hazard ratio function analysis.

RESULTS

The rates of recurrence were significantly higher in patients with ER-negative status for the first two years of follow-up, but not thereafter. Similar results were observed for breast cancer death, and these results held up after adjustment for differences in treatment, age, menopausal status, and tumor burden. When the site of first recurrence was studied, ER-negative status was associated with a significantly higher rate of tumor recurrence in the viscera and soft tissues, while ER-positive status was associated with significantly higher rates of tumor recurrence involving bone.

CONCLUSIONS

The clinical behavior of ER-positive tumors is different from ER-negative cancer. ER status had a pronounced effect on the rates and sites of recurrence. Furthermore, this apparent association diminished over the follow-up period. Recurrence rates were significantly higher in patients with ER-negative status for the first two years of follow-up, but not thereafter.

Prognostic and predictive indicators in operable breast cancer

Because of its biological heterogeneity and wide spectrum of responsiveness to different treatments, breast cancer is a complex disease of difficult clinical management. Over the past several years, knowledge of the molecular mechanisms regulating normal and aberrant cell growth leading to cancer has been enhanced. These advances have enabled the identification of an increasing number of surrogate biomarkers, which have been correlated with prognosis or used as predictors of response to specific treatments. Axillary nodal status, age, tumor size, pathologic grade, and hormone receptor status are the established prognostic and/or predictive factors for selection of adjuvant treatments. The role of new biomarkers, such as p53, HER2/neu, angiogenesis, and the proliferation index value, is promising; however, the clinical value of their determination must be provided by prospective clinical studies.

Factors influencing the prognostic role of oestrogen and progesterone receptor levels in breast cancer--results of the analysis of 670 patients with 11 years of follow-up

In the last two decades, the prognostic role of the steroid hormone receptors has been the subject of a myriad of publications. Nevertheless, its relevance after long-term follow-up is still not clear. The confusion about the prognostic value is mainly due to the difficulty in comparing analyses. Despite different study-designs and statistical approaches, oestrogen (ER) and progesterone (PR) receptors are widely accepted as prognostic factors. Data from 670 breast cancer patients with a median follow-up of 11.4 years were analysed retrospectively. ER and PR were measured by the dextran-coated charcoal (DCC) assay. To investigate the time dependence of the prognostic relevance of ER and PR, separate analyses were done for follow-up shorter and longer than 5 years. Special focus was directed at patients < or =50 and >50 years, node-negative women, in particular those without adjuvant therapy. Univariate and multivariate analyses were performed. In univariate analysis, ER and PR were associated with a significantly longer overall survival at the cut-off levels 10, 20 or 100 fmol/mg protein. The significant survival benefit occurred in the first 5 years of follow-up and remained unchanged in the following period. In the multivariate analyses, only the PR was of significant prognostic value (for PR> or =20 fmol/mg P=0.036, for PR> or =100 P=0.01, Cox analysis). In patients younger than 51 years, only PR was an independent prognosticator at the cut-off level of 100 fmol/mg protein, while in patients >50 years both hormone receptors were not significant. In N0 patients, only the PR reached long-term prognostic independence at a cut-off point of > or =100 fmol/mg (P=0.018). In addition, in the group of node-negative women < or =50 years without adjuvant therapy the PR level reached prognostic significance. The hormone receptor status was a prognostic factor only during the first 5 years of follow-up. Our data suggest that age, lymph node status, length of follow-up and probably the ER/PR assay are important for the evaluation of ER and PR as prognostic variables. In most analyses, PR appeared to be superior to ER in predicting the prognosis of primary breast cancer patients.

A general framework for neural network models on censored survival data

Flexible parametric techniques for regression analysis, such as those based on feed forward artificial neural networks (FFANNs), can be useful for the statistical analysis of censored time data. These techniques are of particular interest for the study of the outcome dependence from several variables measured on a continuous scale, since they allow for the detection of complex non-linear and non-additive effects. Few efforts have been made until now to account for censored times in FFANNs. In the attempt to fill this gap, specific error functions and data representation will be introduced for multilayer perceptron and radial basis function extensions of generalized linear models for survival data.

Modelling cause-specific hazards with radial basis function artificial neural networks: application to 2233 breast cancer patients

The clinical course of a disease is often characterized by the possible occurrence of different types of events acting in a competing way. From a statistical point of view this translates into the need of modelling the dependence of cause-specific hazards as a function of covariates. Generalized linear models with Poisson error have previously been adopted for the analysis of competing risks as a function of discrete covariates. In the present paper an artificial neural network extension for the flexible joint estimation of cause-specific hazards depending on both discrete and continuous covariates is proposed. This approach is based on radial basis function networks which have the advantage of allowing parameter estimation by the adoption of standard software for generalized linear models. We have applied this method to data from 2233 breast cancer patients to investigate the effects of age, tumour size, number of metastatic axillary nodes, histology and tumour site on cause-specific hazards for intra-breast tumour recurrences and distant metastases. The adoption of a radial basis function network made it possible to highlight effects that were not considered by previous analyses of the same data.

Contribution of vascular endothelial growth factor to the Nottingham prognostic index in node-negative breast cancer

The prognostic contribution of intratumour VEGF, the most important factor in tumour-induced angiogenesis, to NPI was evaluated by using flexible modelling in a series of 226 N-primary breast cancer patients in which steroid receptors and cell proliferation were also accounted for. VEGF provided an additional prognostic contribution to NPI mainly within ER-poor tumours.