Thymostimulin versus placebo for palliative treatment of locally advanced or metastasised hepatocellular carcinoma: a phase III clinical trial.

BACKGROUND: Thymostimulin is a thymic peptide fraction with immune-mediated cytotoxicity against hepatocellular carcinoma (HCC) in vitro and palliative efficacy in advanced HCC in two independent phase II trials. The aim of this study was to assess the efficacy of thymostimulin in a phase III trial. METHODS: The study was designed as a prospective randomised, placebo-controlled, double-blind, multicenter clinical phase III trial. Between 10/2002 and 03/2005, 135 patients with locally advanced or metastasised HCC (Karnofsky >or=60%/Child-Pugh <or= 12) were randomised to receive thymostimulin 75 mg s.c. 5x/week or placebo stratified according to liver function. Primary endpoint was twelve-month survival, secondary endpoints overall survival (OS), time to progression (TTP), tumor response, safety and quality of life. A subgroup analysis according to liver function, KPS and tumor stage (Okuda, CLIP and BCLC) formed part of the protocol. RESULTS: Twelve-month survival was 28% [95%CI 17-41; treatment] and 32% [95%CI 19-44; control] with no significant differences in median OS (5.0 [95% CI 3.7-6.3] vs. 5.2 [95% CI 3.5-6.9] months; p = 0.87, HR = 1.04 [95% CI 0.7-1.6]) or TTP (5.3 [95%CI 2.0-8.6] vs. 2.9 [95%CI 2.6-3.1] months; p = 0.60, HR = 1.13 [95% CI 0.7-1.8]). Adjustment for liver function, Karnofsky status or tumor stage did not affect results. While quality of life was similar in both groups, fewer patients on thymostimulin suffered from accumulating ascites and renal failure. CONCLUSIONS: In our phase III trial, we found no evidence of any benefit to thymostimulin in the treatment of advanced HCC and there is therefore no justification for its use as single-agent treatment. The effect of thymostimulin on hepato-renal function requires further confirmation. TRIAL REGISTRATION: Current Controlled Trials ISRCTN64487365.

Pathogenesis of hepatocellular carcinoma.

Hepatocellular carcinoma (HCC) is one of the most common malignant tumours worldwide. The major aetiologies and risk factors for the development of HCC are well defined and some of the multiple steps involved in hepatocarcinogenesis have been elucidated in recent years. However, no clear picture of how and in what sequence these factors interact at the molecular level has emerged yet. Malignant transformation of hepatocytes may occur as a consequence of various aetiologies, such as chronic viral hepatitis, alcohol, and metabolic disorders, in the context of increased cellular turnover induced by chronic liver injury, regeneration and cirrhosis. Activation of cellular oncogenes, inactivation of tumour suppressor genes, genomic instability, including DNA mismatch repair defects and impaired chromosomal segregation, overexpression of growth and angiogenic factors, and telomerase activation may contribute to the development of HCC. Overall, HCCs are genetically very heterogeneous tumours. New technologies, including gene expression profiling and proteomic analyses, should allow us to further elucidate the molecular events underlying HCC development and identify novel diagnostic markers as well as therapeutic targets.

Deep learning via semi-supervised embedding

We show how nonlinear embedding algorithms popular for use with shallow semi-supervised learning techniques such as kernel methods can be applied to deep multilayer architectures, either as a regularizer at the output layer, or on each layer of the architecture. This provides a simple alternative to existing approaches to deep learning whilst yielding competitive error rates compared to those methods, and existing shallow semi-supervised techniques.

Prise en charge du carcinome hépatocettulaire en 2010 [Management of hepatocellular carcinoma]

Hepatocellular carcinoma (HCC) is one of the most frequent malignant tumors worldwide and its incidence has increased over the last years in most developed countries. The majority of HCCs occur in the context of liver cirrhosis. Therefore, patients with cirrhosis and those with hepatitis B virus infection should enter a surveillance program. Detection of a focal liver lesion by ultrasound should be followed by further investigations to confirm the diagnosis and to permit staging. A number of curative and palliative treatment options are available today. The choice of treatment will depend on the tumor stage, liver function and the presence of portal hypertension as well as the general condition of the patient. A multidisciplinary approach is mandatory to offer to each patient the best treatment.

Liver, gastrointestinal and cardiac toxicity in intermediate hepatocellular carcinoma treated with PRECISION TACE with drug-eluting beads: Results from the PRECISION V randomized trial : B-240

Purpose: To evaluate the toxicity focussing on hepatic, gastrointestinal and cardiac parameters following PRECISION TACE with DC Bead? versus conventional transarterial chemoembolization (cTACE) in the treatment of intermediate-stage hepatocellular carcinoma (HCC). Methods and Materials: This prospective, randomized, multicentre study was conducted under best practice trial management and authorized by local institutional review boards. Informed consent was obtained. 212 patients (185 men/27 women; mean: 67 years) were randomized to be treated with DC Beads? or cTACE. The majority of both groups presented in a more advanced stage. Safety was measured by rate of adverse events (South West Oncology Group criteria) and changes in laboratory parameters. Cardiotoxicity was assessed by means of left ventricular ejection fraction (LVEF) in MRI or echocardiography. The results of the two groups were compared using the chi-square test and Student`s t-test. Results: Mean maximum alanine transaminase increase in the DC Bead group was 50% in the cTACE group (p < 0.001) and 59% for aspartate transaminase (p < 0.001). For bilirubin, mean increase was 5.30±15.13 vs. 13.53±73.89 µmol/L. Concerning gastrointestinal disorders, 120 adverse events (AEs) occurred in 57/93 (61.3%) patients in the DC Bead group vs. 114 in 49/108 (45.4%) in cTACE. Concerning hepatobiliary disorders, serious AEs occurred in 8/93 (8.6%) vs. 11/108 (10.2%) patients. LVEF showed an increase in the DC Bead group by +2.7±10.1 percentage points and a small decrease by -1.5±7.6 in the cTACE group, p=0.018. Conclusion: PRECISION TACE is safe, even in more advanced HCC patients. Serious liver and cardiac toxicity were significantly lower in the DC Bead group.

Characteristics of patients with chronic hepatitis C who develop hepatocellular carcinoma.

BACKGROUND AND AIM: Hepatocellular carcinoma (HCC) is the most frequent form of primary liver cancer and chronic infection with hepatitis C virus is one of the main risk factors for HCC. This study analyses the characteristics of the patients with chronic hepatitis C participating in the Swiss Hepatitis C Cohort Study who developed HCC. METHODS: Analysis of the database of the Swiss Hepatitis C Cohort Study, a multicentre study that is being carried out in eight major Swiss hospitals since the year 2000. Patients with chronic hepatitis C and HCC were regrouped and compared to the patients without HCC. RESULTS: Among the 3,390 patients of the cohort, 130 developed an HCC. Age was one of the determining factors. Cirrhosis and its complications ascites and porto-systemic encephalopathy were associated with HCC. Males presented a higher risk for HCC than females. Alcohol consumption was associated with HCC. Diabetes mellitus was an important risk factor, especially in patients with low fibrosis. Patients with Hepatitis C genotype 2 had significantly less HCC than patients with other genotypes. A low socioeconomic status (income, education, profession) was associated with HCC. CONCLUSIONS: Beside the expected characteristics (age, gender, cirrhosis, alcohol), these data stress the role of diabetes mellitus and reveal the importance of low socioeconomic status as a risk factor for HCC in Swiss patients infected with hepatitis C virus. This vulnerable population should be closely monitored.

Multidimensional Adaptations for Open Learning Management Systems

Our work is focused on alleviating the workload for designers of adaptive courses on the complexity task of authoring adaptive learning designs adjusted to specific user characteristics and the user context. We propose an adaptation platform that consists in a set of intelligent agents where each agent carries out an independent adaptation task. The agents apply machine learning techniques to support the user modelling for the adaptation process

Association between the PNPLA3 (rs738409 C>G) variant and hepatocellular carcinoma: Evidence from a meta-analysis of individual participant data.

The incidence of hepatocellular carcinoma (HCC) is increasing in Western countries. Although several clinical factors have been identified, many individuals never develop HCC, suggesting a genetic susceptibility. However, to date, only a few single-nucleotide polymorphisms have been reproducibly shown to be linked to HCC onset. A variant (rs738409 C>G, encoding for p.I148M) in the PNPLA3 gene is associated with liver damage in chronic liver diseases. Interestingly, several studies have reported that the minor rs738409[G] allele is more represented in HCC cases in chronic hepatitis C (CHC) and alcoholic liver disease (ALD). However, a significant association with HCC related to CHC has not been consistently observed, and the strength of the association between rs738409 and HCC remains unclear. We performed a meta-analysis of individual participant data including 2,503 European patients with cirrhosis to assess the association between rs738409 and HCC, particularly in ALD and CHC. We found that rs738409 was strongly associated with overall HCC (odds ratio [OR] per G allele, additive model=1.77; 95% confidence interval [CI]: 1.42-2.19; P=2.78 × 10(-7) ). This association was more pronounced in ALD (OR=2.20; 95% CI: 1.80-2.67; P=4.71 × 10(-15) ) than in CHC patients (OR=1.55; 95% CI: 1.03-2.34; P=3.52 × 10(-2) ). After adjustment for age, sex, and body mass index, the variant remained strongly associated with HCC. Conclusion: Overall, these results suggest that rs738409 exerts a marked influence on hepatocarcinogenesis in patients with cirrhosis of European descent and provide a strong argument for performing further mechanistic studies to better understand the role of PNPLA3 in HCC development.

Trends in mortality from hepatocellular carcinoma in Europe, 1980-2004.

Upward trends in mortality from hepatocellular carcinoma (HCC) were recently reported in the United States and Japan. Comprehensive analyses of most recent data for European countries are not available. Age-standardized (world standard) HCC rates per 100,000 (at all ages, at age 20-44, and age 45-59 years) were computed for 23 European countries over the period 1980-2004 using data from the World Health Organization. Joinpoint regression analysis was used to identify significant changes in trends, and annual percent change were computed. Male overall mortality from HCC increased in Austria, Germany, Switzerland, and other western countries, while it significantly decreased over recent years in countries such as France and Italy, which had large upward trends until the mid-1990s. In the early 2000s, among countries allowing distinction between HCC and other liver cancers, the highest HCC rates in men were in France (6.8/100,000), Italy (6.7), and Switzerland (5.9), whereas the lowest ones were in Norway (1.0), Ireland (0.8), and Sweden (0.7). In women, a slight increase in overall HCC mortality was observed in Spain and Switzerland, while mortality decreased in several other European countries, particularly since the mid-1990s. In the early 2000s, female HCC mortality rates were highest in Italy (1.9/100,000), Switzerland (1.8), and Spain (1.5) and lowest in Greece, Ireland, and Sweden (0.3). In most countries, trends at age 45-59 years were consistent with overall ones, whereas they were more favorable at age 20-44 years in both sexes. CONCLUSION: HCC mortality remains largely variable across Europe. Favorable trends were observed in several European countries mainly over the last decade, particularly in women and in young adults.

New kernel functions and learning methods for text and data mining

Recent advances in machine learning methods enable increasingly the automatic construction of various types of computer assisted methods that have been difficult or laborious to program by human experts. The tasks for which this kind of tools are needed arise in many areas, here especially in the fields of bioinformatics and natural language processing. The machine learning methods may not work satisfactorily if they are not appropriately tailored to the task in question. However, their learning performance can often be improved by taking advantage of deeper insight of the application domain or the learning problem at hand. This thesis considers developing kernel-based learning algorithms incorporating this kind of prior knowledge of the task in question in an advantageous way. Moreover, computationally efficient algorithms for training the learning machines for specific tasks are presented. In the context of kernel-based learning methods, the incorporation of prior knowledge is often done by designing appropriate kernel functions. Another well-known way is to develop cost functions that fit to the task under consideration. For disambiguation tasks in natural language, we develop kernel functions that take account of the positional information and the mutual similarities of words. It is shown that the use of this information significantly improves the disambiguation performance of the learning machine. Further, we design a new cost function that is better suitable for the task of information retrieval and for more general ranking problems than the cost functions designed for regression and classification. We also consider other applications of the kernel-based learning algorithms such as text categorization, and pattern recognition in differential display. We develop computationally efficient algorithms for training the considered learning machines with the proposed kernel functions. We also design a fast cross-validation algorithm for regularized least-squares type of learning algorithm. Further, an efficient version of the regularized least-squares algorithm that can be used together with the new cost function for preference learning and ranking tasks is proposed. In summary, we demonstrate that the incorporation of prior knowledge is possible and beneficial, and novel advanced kernels and cost functions can be used in algorithms efficiently.

Dysregulation of apoptosis in hepatocellular carcinoma cells

Hepatocellular carcinoma (HCC) is a major health problem, being the sixth most common cancer world-wide. Dysregulation of the balance between proliferation and cell death represents a pro-tumorigenic principle in human hepatocarcinogenesis. This review updates the recent relevant contributions reporting molecular alterations for HCC that induce an imbalance in the regulation of apoptosis. Alterations in the expression and/or activation of p53 are frequent in HCC cells, which confer on them resistance to chemotherapeutic drugs. Many HCCs are also insensitive to apoptosis induced either by death receptor ligands, such as FasL or TRAIL, or by transforming growth factor-beta (TGF-beta). Although the expression of some pro-apoptotic genes is decreased, the balance between death and survival is dysregulated in HCC mainly due to overactivation of anti-apoptotic pathways. Indeed, some molecules involved in counteracting apoptosis, such as Bcl-XL, Mcl-1, c-IAP1, XIAP or survivin are over-expressed in HCC cells. Furthermore, some growth factors that mediate cell survival are up-regulated in HCC, as well as the molecules involved in the machinery responsible for cleavage of their pro-forms to an active peptide. The expression and/or activation of the JAK/STAT, PI3K/AKT and RAS/ERKs pathways are enhanced in many HCC cells, conferring on them resistance to apoptotic stimuli. Finally, recent evidence indicates that inflammatory processes, as well as the epithelial-mesenchymal transitions that occur in HCC cells to facilitate their dissemination, are related to cell survival. Therefore, therapeutic strategies to selectively inhibit anti-apoptotic signals in liver tumor cells have the potential to provide powerful tools to treat HCC.

A Sparse Bayesian Approach for Joint Feature Selection and Classifier Learning

Peer-reviewed

Design and evaluation of learning algorithms for dynamic resource management in virtual networks

Network virtualisation is considerably gaining attentionas a solution to ossification of the Internet. However, thesuccess of network virtualisation will depend in part on how efficientlythe virtual networks utilise substrate network resources.In this paper, we propose a machine learning-based approachto virtual network resource management. We propose to modelthe substrate network as a decentralised system and introducea learning algorithm in each substrate node and substrate link,providing self-organization capabilities. We propose a multiagentlearning algorithm that carries out the substrate network resourcemanagement in a coordinated and decentralised way. The taskof these agents is to use evaluative feedback to learn an optimalpolicy so as to dynamically allocate network resources to virtualnodes and links. The agents ensure that while the virtual networkshave the resources they need at any given time, only the requiredresources are reserved for this purpose. Simulations show thatour dynamic approach significantly improves the virtual networkacceptance ratio and the maximum number of accepted virtualnetwork requests at any time while ensuring that virtual networkquality of service requirements such as packet drop rate andvirtual link delay are not affected.

Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.