[show abstract][hide abstract] ABSTRACT: in primary breast cancers dichotomic classification of E-cadherin expression, according to an arbitrary cutoff, may be inadequate and lead to loss of prognostic significance or contrasting prognostic indications. We aimed to assess the prognostic value of high and low E-cadherin levels in a consecutive case series (204 cases) of unilateral node-negative non-lobular breast cancer patients with a 8-year median follow-up and that did not receive any adjuvant therapy after surgery.
expression of E-cadherin was investigated by immunohistochemistry and assessed according to conventional score (0, 1+, 2+, 3+). Multiple correspondence analysis was used to visualise associations of both categorical and continuous variables. The impact of E-cadherin expression on patients outcome was evaluated in terms of event-free survival curves by the Kaplan-Meier method and proportional hazard Cox model.
respect to intermediate E-cadherin expression values (2+), high (3+) or low (0 to 1+) E-cadherin expression levels had a negative prognostic impact. In fact, both patients with a low-to-nil (score 0 to 1+) expression level of E-cadherin and patients with a high E-cadherin expression level (score 3+) demonstrated an increased risk of failure (respectively, hazard ratio (HR)=1.71, confidence interval (CI)=0.72-4.06 and HR=4.22, CI=1.406-12.66) and an interesting association with young age.
the findings support the evidence that high expression values of E-cadherin are not predictive for a good prognosis and may help to explain conflicting evidence on the prognostic impact of E-cadherin in breast cancer when assessed on dichotomic basis.
British Journal of Cancer 11/2010; 103(12):1835-9. · 5.08 Impact Factor
[show abstract][hide abstract] ABSTRACT: Pharmaceutical companies face twin pressures of improving the throughput of new products and responding to increased requirements by regulatory bodies for risk evaluation and mitigation both pre- and post-marketing. This paper gives a brief overview of flexible outcome models to improve the accuracy of these studies. The focus is on the discovery of niche populations in secondary analysis and monitoring of adverse events in observational studies of new products. Analytical results illustrate the power of the methods and they can be mapped onto Boolean filters to represent populations with particular benefit/risk ratios.
Neural Networks (IJCNN), The 2010 International Joint Conference on; 08/2010
[show abstract][hide abstract] ABSTRACT: This article is not intended as a comprehensive survey of data mining applications in cancer. Rather, it provides starting points for further, more targeted, literature searches, by embarking on a guided tour of computational intelligence applications in cancer medicine, structured in increasing order of the physical scales of biological processes.
[show abstract][hide abstract] ABSTRACT: Time-to-event analysis is important in a wide range of applications from clinical prognosis to risk modeling for credit scoring and insurance. In risk modeling, it is sometimes required to make a simultaneous assessment of the hazard arising from two or more mutually exclusive factors. This paper applies to an existing neural network model for competing risks (PLANNCR), a Bayesian regularization with the standard approximation of the evidence to implement automatic relevance determination (PLANNCR-ARD). The theoretical framework for the model is described and its application is illustrated with reference to local and distal recurrence of breast cancer, using the data set of Veronesi (1995).
IEEE Transactions on Neural Networks 10/2009; · 2.95 Impact Factor
[show abstract][hide abstract] ABSTRACT: Clustering algorithms will, in general, either partition a given data set into a pre-specified number of clusters or will produce a hierarchy of clusters. In this paper we analyse several different clustering techniques and apply them to a particular data set of breast cancer data. When we do not know a priori which is the best number of groups, we use a range of different validity indices to test the quality of clustering results and to determine the best number of clusters. While for the K-means method there is not absolute agreement among the indices as to which is the best number of clusters, for the PAM algorithm all the indices indicate 4 as the best cluster number.
Advances in Medical, Signal and Information Processing, 2008. MEDSIP 2008. 4th IET International Conference on; 08/2008
[show abstract][hide abstract] ABSTRACT: Accurate modelling of time-to-event data is of particular importance for both exploratory and predictive analysis in cancer, and can have a direct impact on clinical care. This study presents a detailed double-blind evaluation of the accuracy in out-of-sample prediction of mortality from two generic non-linear models, using artificial neural networks benchmarked against a partial logistic spline, log-normal and COX regression models. A data set containing 2880 samples was shared over the Internet using a purpose-built secure environment called GEOCONDA (www.geoconda.com). The evaluation was carried out in three parts. The first was a comparison between the predicted survival estimates for each of the four survival groups defined by the TNM staging system, against the empirical estimates derived by the Kaplan-Meier method. The second approach focused on the accurate prediction of survival over time, quantified with the time dependent C index (C(td)). Finally, calibration plots were obtained over the range of follow-up and tested using a generalization of the Hosmer-Lemeshow test. All models showed satisfactory performance, with values of C(td) of about 0.7. None of the models showed a systematic tendency towards over/under estimation of the observed survival at tau=3 and 5 years. At tau=10 years, all models underestimated the observed survival, except for COX regression which returned an overestimate. The study presents a robust and unbiased benchmarking methodology using a bespoke web facility. It was concluded that powerful, recent flexible modelling algorithms show a comparative predictive performance to that of more established methods from the medical and biological literature, for the reference data set.
Computers in Biology and Medicine 09/2007; 37(8):1108-20. · 1.16 Impact Factor
[show abstract][hide abstract] ABSTRACT: This paper presents a Bayesian Neural Network for the analysis of Competing Risk (CR) data model. Based on a previously developed non-linear model namely Partial Logistic Artificial Neural Network (PLANN) with Automatic Relevance Determination (ARD), this paper proposes an extension for the flexible joint estimation of cause-specific hazards depending on both discrete and continuous covariates (PLANN-CR-ARD) and for censored data. The Bayesia analysis uses Gaussian priors for the neural network parameters and the likelihood function based on the competing risk data is identified as the cross-entropy function. The PLANN-CR-ARD model is illustrated with analyses of an Intra-Ocular Melanoma dataset and comparison with the non-parametric Nelson-Allen estimates of the cause-specific cumulative hazards functions.
Advances in Medical, Signal and Information Processing, 2006. MEDSIP 2006. IET 3rd International Conference On; 08/2006
[show abstract][hide abstract] ABSTRACT: 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.
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.
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.
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.
Annals of Oncology 11/2003; 14(10):1484-93. · 7.38 Impact Factor
[show abstract][hide abstract] ABSTRACT: 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.
British Journal of Cancer 08/2003; 89(2):268-70. · 5.08 Impact Factor
[show abstract][hide abstract] ABSTRACT: The aims of the present investigation were to evaluate the association between serum CA15.3 levels and other biological and clinical variables and its prognostic role in patients with node-negative breast cancer. We evaluated 362 patients operated upon primary breast cancer from 1982 to 1992 (median follow-up 69 months). Serum CA15.3 was measured by an immunoradiometric assay. The association between variables was investigated by a Principal Component Analysis (PCA) and the prognostic role of CA15.3 on relapse-free survival (RFS) was investigated by Cox regression models adjusting for age, oestrogen receptor (ER), tumour stage, and ER x age interaction, with both the likelihood ratio test and Harrell's c statistic. The prognostic contribution of CA 15.3 was highly significant. Log relative hazard of relapse was constant until approximately 10 (U/ml) of CA15.3 and increased thereafter with increasing marker levels. CA15.3 showed a significant contribution using as a cut-off point a value of 31 U/ml. However, the contribution to the model of the marker as a continuous variable is much greater. From these findings, we can conclude that: (i) CA15.3 is a prognostic marker in node-negative breast cancer; (ii) its relationship with prognosis is continuous, with the risk of relapse increasing progressively from approximately 10 U/ml.
European Journal of Cancer 07/2002; 38(9):1181-8. · 5.06 Impact Factor
[show abstract][hide abstract] ABSTRACT: 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.
Statistics in Medicine 01/2002; 20(24):3677-94. · 2.04 Impact Factor
[show abstract][hide abstract] ABSTRACT: 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.
British Journal of Cancer 10/2001; 85(6):795-7. · 5.08 Impact Factor
[show abstract][hide abstract] ABSTRACT: There is considerable interest in biologic markers able to predict the response of cancer patients to therapy. HER2 overexpression is a potential indicator of responsiveness to doxorubicin and paclitaxel and of unresponsiveness to tamoxifen in breast carcinoma patients. However, the significance of HER2 overexpression in responsiveness to cyclophosphamide, methotrexate, and fluorouracil (CMF) has remained unclear. In this study, we investigated this issue in the 386 breast cancer patients in the first CMF controlled clinical trial with a 20-year follow-up.
Node-positive breast carcinoma patients were randomly assigned to receive either no further treatment after radical mastectomy (179 women) or 12 monthly cycles of adjuvant CMF chemotherapy (207 women). Overexpression of HER2 and the status of other tumor variables was assessed by immunohistochemistry in at least 324 (84%) of the 386 patients. Statistical analyses were performed to assess the efficacy of CMF treatment for the subgroups defined by HER2 and the status of other variables using a Bayesian approach. The end points considered were relapse-free survival (RFS) and cause-specific survival (CSS).
Bayesian analysis of the treatment effect for HER2 and other variables indicated a clinical benefit from CMF treatment in all subgroups defined according to variables status. In particular regarding HER2 status, Bayesian estimates of RFS hazard ratios were equal to 0.484 and 0.641 and estimates of CSS hazard ratios were equal to 0.495 and 0.730 for HER2-positive and -negative tumors, respectively.
CMF treatment showed a clinical benefit in the considered subgroups, defined according to HER2 and other tumor variables status. Patients with HER2-positive or HER2-negative tumors benefit from CMF treatment, and the poor prognosis associated with the HER2 overexpression in the untreated group could be completely overcome by the chemotherapy treatment.
Journal of Clinical Oncology 02/2001; 19(2):329-35. · 18.04 Impact Factor
[show abstract][hide abstract] ABSTRACT: Thrombospondins (TSP(s)) are a multigene family of five secreted glycoproteins involved in the regulation of cell proliferation, adhesion and migration. Two members of the TSP family, namely TSP-1 and TSP-2, are also naturally occurring inhibitors of angiogenesis. The aim of the present study was to determine the prognostic significance of the determination of TSP-1 and -2 and their correlation with the angiogenic peptides vascular endothelial growth factor (VEGF) and thymidine phosphorylase (TP), as well as with other biological and clinicopathological features investigated.
We evaluated a series of 168 women with node-negative breast cancer with a median follow-up period of 66 months, not treated with adjuvant therapy. The cytosolic levels of TSP-1 and -2 were determined in the primary tumour by a commercially available immunometric assay.
We found that 166 tested tumours had measurable levels of TSP-1 and -2 protein (median value 5.978, range 0.579-31.410 ng/mg of protein). On the basis of Spearman's rank correlation coefficient, a weak inverse association of TSP-1 and -2 with tumour size and cathepsin D was found. Moreover, principal component analysis on ranks evidenced a poor association between TSP-1 and -2, VEGF and TP. The results of the clinical outcome were analysed by both univariate and multivariate [for relapse-free survival (RFS) only]) Cox regression models. TSP-1 and -2 were not significant prognostic factors in univariate analysis for either RFS (p = 0.427) or overall survival (p = 0.069). To investigate the 'angiogenic balance hypothesis', bivariate analyses were performed to investigate the interactions of TSP-1 and -2 with VEGF, TP or p53, but none were included in the selected models. Finally, in multivariate analysis for RFS a baseline model, previously defined in a larger case series and inclusive of VEGF, TP and their interaction was adopted. It was highly significant (p = 0.002, Harrell c statistic value of 0.703); but when TSP-1 and -2 were added, their contribution was negligible (p = 0.731, Harrell c statistic value of 0.705).
The results of this study suggest that TSP-1 and -2 do not provide additional prognostic contribution to the joint effects of VEGF and TP. In the series of node-negative breast cancer patients investigated, determination of the angiogenic peptides VEGF and TP gave significant prognostic information. On the contrary, TSP-1 and -2, potential naturally occurring negative regulators of angiogenesis, lacked prognostic value.
[show abstract][hide abstract] ABSTRACT: The frog embryo teratogenesis assay-Xenopus (FETAX) is a powerful and flexible bioassay that makes use of the embryos of the anuran Xenopus laevis. FETAX satisfies the requirements of low cost, reliability and reproducibility and, thanks to its three endpoints (i.e., mortality, teratogenicity and growth inhibition) can detect the xenobiotics that affect embryonic development. In this paper, we have used FETAX to evaluate samples of soils collected in an oil-contaminated area. Embryos were exposed directly to the soil to be tested. Particular attention was devoted to provide a statistical procedure for analysing mortality and malformation data as well as growth retardation.
[show abstract][hide abstract] ABSTRACT: Non-palpable breast cancers are often in situ or smaller and have less nodal and distant metastases than palpable lesions. They represent a heterogeneous group of tumours, which may have different prognostic behaviour. We analysed a retrospective series of 982 non-palpable breast cancers assessed histologically at the National Cancer Institute of Milan from 1985 to 1995, following pre-operative mammography-guided localization. The association between mammographic data (parenchymal pattern and findings), patient age and tumour histology was investigated by review of clinical records and statistical modelling. We also investigated the association between the presence or absence of microcalcification as a mammographic finding and pathological tumour characteristics (tumour size, axillary nodes status and grading) or receptor status for oestrogen (ER) and progesterone (PgR). In situ disease or invasive tumour with an intraductal component, whether extensive or not, were commoner in young women and mammography more frequently showed a dense parenchymal pattern and microcalcifications in these cases. In older women (55 years or more), a fatty breast pattern, nodular opacities with or without microcalcifications, and invasive tumours of the ductal, lobular, mixed or other types were closely related. When the relationships between mammographic findings, pathological tumour characteristics and receptor status were investigated for invasive cancers, there was an association between the presence of microcalcifications and less favourable tumour characteristics.
British Journal of Radiology 08/2000; 73(871):698-705. · 1.22 Impact Factor