Re-induction using whole cell melanoma vaccine genetically modified to melanoma stem cells-like beyond recurrence extends long term survival of high risk resected patients - updated results

BACKGROUND

AGI-101H is an allogeneic gene modified whole cell therapeutic melanoma vaccine, evaluated in over 400 melanoma patients in the adjuvant and therapeutic settings. We present updated long-term survival results from two single-arm, phase II adjuvant trials (Trial 3 and Trial 5) with the focus on treatment beyond recurrence of the disease.

METHODS

Patients with resected high-risk melanoma (stage IIIB-IV) were enrolled to Trial 3 (n = 99) and Trial 5 (n = 97). The primary endpoint was disease-free survival (DFS), and the secondary was overall survival (OS). In the induction phase, the vaccine was administered every 2 weeks (eight times), followed by the maintenance phase every month until progression. At progression, maintenance was continued or re-induction was applied with or without surgery.

RESULTS

In Trial 3, the 10-year DFS was equal to 33.0% overall and to 52.4, 25.0, and 8.7% for stage IIIB, IIIC, and stage IV patients, respectively. In Trial 5, the overall 10-year DFS was equal to 24.2%, and to 37.5, 18.0, and 17.6% for stage IIIB, IIIC, and stage IV patients, respectively. In Trial 3, the 10-year OS was equal to 42.3% overall, and to 59.5, 37.5, and 17.4% for stage IIIB, IIIC, and stage IV patients, respectively. In Trial 5, the 10-year OS was equal to 34.3% overall and to 46.9, 28.0, and 29.4% for stage IIIB, IIIC, and stage IV patients, respectively. Among the 65 patients of Trial 3 who developed progression, 43 received re-induction with (n = 22) or without (n = 21) surgery. Two patients received surgery without re-induction. All the 22 progressing patients, who did not receive re-induction, died. Among the 75 patients of Trial 5 who experienced progression, 39 received re-induction with (n = 21) or without (n = 18) surgery. Among the 36 progressing patients who did not receive the re-induction, 35 died. Surgery and re-induction reduced (independently) the increase of mortality after progression in both trials, with the effect of re-induction reaching statistical significance in Trial 5.

CONCLUSIONS

Vaccination beyond recurrence of the disease with additional re-induction combined with surgery or alone increased long term survival of melanoma patients. However, further studies on larger patient cohorts are required.

TRIAL REGISTRATION

Central Evidence of Clinical Trials (EudraCT Number 2008-003373-40 ).

Surrogate endpoints in advanced sarcoma trials: a meta-analysis

Background

Alternative endpoints to overall survival (OS) are frequently used to assess treatment efficacy in randomized controlled trials (RCT). Their properties in terms of surrogate outcomes for OS need to be assessed. We evaluated the surrogate properties of progression-free survival (PFS), time-to-progression (TTP) and time-to-treatment failure (TTF) in advanced soft tissue sarcomas (STS).

Results

A total of 21 trials originally met the selection criteria and 14 RCTs (N = 2846) were included in the analysis. Individual-level associations were moderate (highest for 12-month PFS: Spearman’s rho = 0.66; 95% CI [0.63; 0.68]). Trial-level associations were ranked as low for the three endpoints as per the IQWiG criterion.

Materials and Methods

We performed a meta-analysis using individual-patient data (IPD). Phase II/III RCTs evaluating therapies for adults with advanced STS were eligible. We estimated the individual- and the trial-level associations between then candidate surrogates and OS. Statistical methods included weighted linear regression and the two-stage model introduced by Buyse and Burzykowski. The strength of the trial-level association was ranked according to the German Institute for Quality and Efficiency in Health Care (IQWiG) guidelines.

Conclusions

Our results do not support strong surrogate properties of PFS, TTP and TTF for OS in advanced STS.

Abstract LB-373: Comprehensive analysis of cancer stemness

Cancer progression involves the gradual loss of a differentiated phenotype and acquisition of progenitor and stem cell-like features. Here, we provide new stemness indices for assessing the degree of oncogenic dedifferentiation. We took advantage of an innovative one-class logistic regression machine learning algorithm (OCLR) to extract transcriptomic and epigenetic feature sets derived from non-transformed pluripotent stem cells and their differentiated progenies. Using OCLR, we were able to sort TCGA tumor samples by stemness phenotype and identify previously undiscovered biological mechanisms associated with the dedifferentiated oncogenic state. Analyses of tumor microenvironment revealed the correlation of cancer stemness with immune checkpoint expression and infiltrating immune system cells not previously anticipated. We have shown the de-differentiated oncogenic phenotype increased in the metastatic tumor that further justify their more aggressive phenotype. Application of our stemness indices reveals features of intra-tumor heterogeneity in molecular profiles obtained from the single-cell analyses. Finally, the machine learning-based indices allowed for the identification of chemical compounds and novel targets for the cancer therapies aiming at tumor differentiation. Our findings provide new prognostic signatures that enable cancer biologists and oncologists to quantify the impact of tumor stemness on outcome across cancer types and may help to pave the way for progress in treatment strategies for cancer patients.

Citation Format: Tathiane M. Malta, Artem Sokolov, Andrew J. Gentles, Tomasz Burzykowski, Laila Poisson, John Weinstein, Bożena Kamińska, Joerg Huelsken, Larsson Omberg, Olivier Gevaert, Antonio Colaprico, Patrycja Czerwińska, Sylwia Mazurek, Lopa Mishra, Holger Heyn, Alex Krasnitz, Andrew K. Godwin, Alexander J. Lazar, The Cancer Genome Atlas Research Network, Joshua M. Stuart, Katherine Hoadley, Peter W. Laird, Houtan Noushmehr, Maciej Wiznerowicz. Comprehensive analysis of cancer stemness [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-373.

Computational methods and challenges in hydrogen/deuterium exchange mass spectrometry

Hydrogen/Deuterium exchange (HDX) has been applied, since the 1930s, as an analytical tool to study the structure and dynamics of (small) biomolecules. The popularity of using HDX to study proteins increased drastically in the last two decades due to the successful combination with mass spectrometry (MS). Together with this growth in popularity, several technological advances have been made, such as improved quenching and fragmentation. As a consequence of these experimental improvements and the increased use of protein-HDXMS, large amounts of complex data are generated, which require appropriate analysis. Computational analysis of HDXMS requires several steps. A typical workflow for proteins consists of identification of (non-)deuterated peptides or fragments of the protein under study (local analysis), or identification of the deuterated protein as a whole (global analysis); determination of the deuteration level; estimation of the protection extent or exchange rates of the labile backbone amide hydrogen atoms; and a statistically sound interpretation of the estimated protection extent or exchange rates. Several algorithms, specifically designed for HDX analysis, have been proposed. They range from procedures that focus on one specific step in the analysis of HDX data to complete HDX workflow analysis tools. In this review, we provide an overview of the computational methods and discuss outstanding challenges. © 2016 Wiley Periodicals, Inc. Mass Spec Rev 36:649-667, 2017.

A Poisson approach to the validation of failure time surrogate endpoints in individual patient data meta-analyses

Surrogate endpoints are often used in clinical trials instead of well-established hard endpoints for practical convenience. The meta-analytic approach relies on two measures of surrogacy: one at the individual level and one at the trial level. In the survival data setting, a two-step model based on copulas is commonly used. We present a new approach which employs a bivariate survival model with an individual random effect shared between the two endpoints and correlated treatment-by-trial interactions. We fit this model using auxiliary mixed Poisson models. We study via simulations the operating characteristics of this mixed Poisson approach as compared to the two-step copula approach. We illustrate the application of the methods on two individual patient data meta-analyses in gastric cancer, in the advanced setting (4069 patients from 20 randomized trials) and in the adjuvant setting (3288 patients from 14 randomized trials).

Abstract LB-004: Molecular hallmarks of cancer: Stemness

Stemness, defined as the potential for self-renewal and de-differentiation from the cell-of-origin, has been initially attributed to normal stem cells that possess the ability to give rise to all cell types in the adult organism. Cancer stem cells have been identified in most if not all hematological malignancies as well as in solid tumors. CSCs are postulated to be amongst the most resistant cells to various forms of chemotherapy and radiotherapy and contribute to relapse and metastasis and hence inferior clinical outcomes.

Here, we have performed a comprehensive and multi-platform analyses of the stemness features in 33 cancer types totalling more than 10,000 mostly primary, but also some metastatic and recurrent samples. In the first step, we have derived signatures to measure the stemness using molecular profiles of normal cells with various degrees of stemness from publicly available datasets. By multiplatform analyses of the transcriptome, methylome, and chromatin markers using different machine learning computational approaches, we have obtained four independent stemness scores. Initial validation revealed comparable performance of these computational metrics for sorting the TCGA samples. The cancer types with previously documented features of de-differentiation have higher stemness score in contrast to more differentiated cancer types. More detailed analyses of recently published molecular subtypes of gliomas have revealed a strong association of high stemness scores with the worst clinical outcome, further confirming our initial hypotheses.

By adopting machine learning algorithms, we have counted percentage of immune cells of both innate and acquired response infiltrating tumor tissue. Further deconvolution of the TCGA cancers allowed quantification of the tumour microenvironment composed of cancer-associated fibroblasts and the endothelial cells. Integration of gene expression and DNA methylation datasets defines classic hallmarks of pluripotent signatures which shed new light into biological processes regulating and maintaining oncogenic dedifferentiation.

Our ongoing analyses of the TCGA samples sorted by the obtained metrics involve: (1) activity of the selected hallmarks of cancer that have been shown as features of the CSCs; (2) correlation with the tumor pathology grading and clinical outcomes; (3) identification of potential drivers for development and/or repositioning of drugs targeting tumor self-renewal and de-differentiation potential; and (4) development of novel biomarkers for selection of patients that will respond to these novel therapies.

In summary, we defined molecular signatures of cancer stem cells that enabled classification of TCGA cancer types and identification of tumors with stem/progenitor-like phenotypes, associated with poor clinical outcomes. Understanding stemness-like hallmark of cancer will pave the way for novel diagnostics and therapies for cancer patients.

Citation Format: Tathiane Malta, Artem Sokolov, Andrew J. Gentles, Tomasz Burzykowski, Olivier Gevaert, The Stemness Analysis Working Group TCGA PanCancerNetwork, Peter W. Laird, Houtan Noushmehr, Maciej Wiznerowicz. Molecular hallmarks of cancer: Stemness [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr LB-004. doi:10.1158/1538-7445.AM2017-LB-004

Tumor-size-based endpoints as surrogates for overall survival in the ARCAD Advanced Colorectal Cancer Database

766

Background: Tumor-size-based endpoints such as depth of response have recently become a focus of investigations related to search for early on-treatment predictive markers for patients with colorectal cancer (Piessevaux et al. 2013; Cremolini et al., 2015; Heinemann et al., 2015). We evaluate the use of time-to-nadir (TTN) and depth-of-nadir (DoN) as potential surrogates for overall survival (OS). Methods: We have used the data from trials included in the ARCAD Advanced Colorectal Cancer Database. We have considered three sets of treatment comparisons: comparisons involving only chemotherapy agents (CT); involving anti-angiogenesic agents (ANG); and involving anti-EGFR agents (EGFR). For each of the sets separately, we applied a two-stage modelling approach (Renard et al. 2002) to jointly analyze the relative change of tumor size versus baseline (RCTS) and OS. In particular, in the first stage, a joint model for the repeated measurements of RCTS and OS was fitted to the data. Based on the model, treatment effects on OS and RCTS were estimated. Treatment effects on OS were expressed as log-hazard-ratios, while the effects on RCTS were expressed in terms of differences in the time-to-nadir (TTN) and in the depth of nadir (DoN). In the second stage, a linear regression was fitted through the estimated treatment effects on TTN/DoN and OS. The coefficient of determination (R2), computed by using the comparison-specific sample-size as weights, was used to quantify the strength of association between the treatment effects on TTN/DoN and OS. In particular, values of R2 larger than 0.75 were considered as indicating a good surrogate. Results: The CT set included 18 comparisons with 5726 pts. in total. For the treatment effects on TTN and OS, the value of R2 was estimated to be equal to 0.31, while for DoN and OS, it was equal to 0.01. For the ANG set (11 comparisons, 3964 pts.) the values of R2 were equal to 0.26 and 0.21, respectively, while for the EGFR set (16 comparisons, 4687 pts.) they were equal to 0.10 and 0.45, respectively. Conclusions: These preliminary results suggest that TTN and DoN are not good surrogates for OS in advanced colorectal cancer. More detailed analyses are ongoing and will be presented at the meeting.

Estimation of diagnostic accuracy of a combination of continuous biomarkers allowing for conditional dependence between the biomarkers and the imperfect reference-test

Estimating biomarker-index accuracy when only imperfect reference-test information is available is usually performed under the assumption of conditional independence between the biomarker and imperfect reference-test values. We propose to define a latent normally-distributed tolerance-variable underlying the observed dichotomous imperfect reference-test results. Subsequently, we construct a Bayesian latent-class model based on the joint multivariate normal distribution of the latent tolerance and biomarker values, conditional on latent true disease status, which allows accounting for conditional dependence. The accuracy of the continuous biomarker-index is quantified by the AUC of the optimal linear biomarker-combination. Model performance is evaluated by using a simulation study and two sets of data of Alzheimer's disease patients (one from the memory-clinic-based Amsterdam Dementia Cohort and one from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database). Simulation results indicate adequate model performance and bias in estimates of the diagnostic-accuracy measures when the assumption of conditional independence is used when, in fact, it is incorrect. In the considered case studies, conditional dependence between some of the biomarkers and the imperfect reference-test is detected. However, making the conditional independence assumption does not lead to any marked differences in the estimates of diagnostic accuracy.

Normalisation of brain spectroscopy findings in Niemann-Pick disease type C patients treated with miglustat

Niemann-Pick disease type C (NP-C) is a fatal progressive neurolipidosis involving neuronal storage of cholesterol and gangliosides. Miglustat, an inhibitor of glycosphingolipid synthesis, has been approved to treat neurological manifestations in adults and children with NP-C. This open-label observational study in adults with confirmed NP-C evaluated the efficacy of miglustat (200 mg t.i.d.) based on composite functional disability (CFD) scores and brain proton magnetic resonance spectroscopy (H-MRS) measurement of choline (Cho)/N-acetyl aspartate (NAA) ratio in the centrum ovale. Overall, 16 patients were included and received miglustat for a mean period of 30.6 months: 12 continued on miglustat throughout follow up, and 4 discontinued miglustat because of adverse effects (n = 2) or perceived lack of efficacy (n = 2). In the 'continued' subgroup, the mean (SD) annual progression of CFD scores decreased from 0.75 (0.94) before treatment to 0.29 (1.29) during the period between miglustat initiation and last follow-up. In the discontinued subgroup, CFD progression increased from 0.48 (0.44) pre-treatment to 1.49 (1.31) at last follow up (off treatment). Mean (SD) Cho/NAA ratio [normal level 0.48 (0.076)] decreased during miglustat treatment in the continued subgroup: 0.64 (0.12) at baseline (miglustat initiation), 0.59 (0.17) at 12-month follow up, and 0.48 (0.09) at 24-month follow up. Cho/NAA ratio remained relatively stable in the discontinued subgroup: 0.57 (0.15), 0.53 (0.04) and 0.55 (0.09), respectively. In conclusion, H-MRS Cho/NAA ratio might serve as an objective, quantitative neurological marker of brain dysfunction in NP-C, allowing longitudinal analysis of the therapeutic effect of miglustat.

The Detection of Metabolite-Mediated Gene Module Co-Expression Using Multivariate Linear Models

Investigating whether metabolites regulate the co-expression of a predefined gene module is one of the relevant questions posed in the integrative analysis of metabolomic and transcriptomic data. This article concerns the integrative analysis of the two high-dimensional datasets by means of multivariate models and statistical tests for the dependence between metabolites and the co-expression of a gene module. The general linear model (GLM) for correlated data that we propose models the dependence between adjusted gene expression values through a block-diagonal variance-covariance structure formed by metabolic-subset specific general variance-covariance blocks. Performance of statistical tests for the inference of conditional co-expression are evaluated through a simulation study. The proposed methodology is applied to the gene expression data of the previously characterized lipid-leukocyte module. Our results show that the GLM approach improves on a previous approach by being less prone to the detection of spurious conditional co-expression.

Development of a diagnostic test based on multiple continuous biomarkers with an imperfect reference test

Ignoring the fact that the reference test used to establish the discriminative properties of a combination of diagnostic biomarkers is imperfect can lead to a biased estimate of the diagnostic accuracy of the combination. In this paper, we propose a Bayesian latent-class mixture model to select a combination of biomarkers that maximizes the area under the ROC curve (AUC), while taking into account the imperfect nature of the reference test. In particular, a method for specification of the prior for the mixture component parameters is developed that allows controlling the amount of prior information provided for the AUC. The properties of the model are evaluated by using a simulation study and an application to real data from Alzheimer's disease research. In the simulation study, 100 data sets are simulated for sample sizes ranging from 100 to 600 observations, with a varying correlation between biomarkers. The inclusion of an informative as well as a flat prior for the diagnostic accuracy of the reference test is investigated. In the real-data application, the proposed model was compared with the generally used logistic-regression model that ignores the imperfectness of the reference test. Conditional on the selected sample size and prior distributions, the simulation study results indicate satisfactory performance of the model-based estimates. In particular, the obtained average estimates for all parameters are close to the true values. For the real-data application, AUC estimates for the proposed model are substantially higher than those from the 'traditional' logistic-regression model.

An information-theoretic approach for the evaluation of surrogate endpoints based on causal inference

In this work a new metric of surrogacy, the so-called individual causal association (ICA), is introduced using information-theoretic concepts and a causal inference model for a binary surrogate and true endpoint. The ICA has a simple and appealing interpretation in terms of uncertainty reduction and, in some scenarios, it seems to provide a more coherent assessment of the validity of a surrogate than existing measures. The identifiability issues are tackled using a two-step procedure. In the first step, the region of the parametric space of the distribution of the potential outcomes, compatible with the data at hand, is geometrically characterized. Further, in a second step, a Monte Carlo approach is proposed to study the behavior of the ICA on the previous region. The method is illustrated using data from the Collaborative Initial Glaucoma Treatment Study. A newly developed and user-friendly R package Surrogate is provided to carry out the evaluation exercise.

Statistical monitoring of data quality and consistency in the Stomach Cancer Adjuvant Multi-institutional Trial Group Trial

INTRODUCTION

Data quality may impact the outcome of clinical trials; hence, there is a need to implement quality control strategies for the data collected. Traditional approaches to quality control have primarily used source data verification during on-site monitoring visits, but these approaches are hugely expensive as well as ineffective. There is growing interest in central statistical monitoring (CSM) as an effective way to ensure data quality and consistency in multicenter clinical trials.

METHODS

CSM with SMART™ uses advanced statistical tools that help identify centers with atypical data patterns which might be the sign of an underlying quality issue. This approach was used to assess the quality and consistency of the data collected in the Stomach Cancer Adjuvant Multi-institutional Trial Group Trial, involving 1495 patients across 232 centers in Japan.

RESULTS

In the Stomach Cancer Adjuvant Multi-institutional Trial Group Trial, very few atypical data patterns were found among the participating centers, and none of these patterns were deemed to be related to a quality issue that could significantly affect the outcome of the trial.

DISCUSSION

CSM can be used to provide a check of the quality of the data from completed multicenter clinical trials before analysis, publication, and submission of the results to regulatory agencies. It can also form the basis of a risk-based monitoring strategy in ongoing multicenter trials. CSM aims at improving data quality in clinical trials while also reducing monitoring costs.

Differences in the Elemental Isotope Definition May Lead to Errors in Modern Mass-Spectrometry-Based Proteomics

The elemental isotope definition used to calculate the theoretical masses and isotope distribution of (bio)molecules is considered to be a fixed, universal standard in mass-spectrometry-based proteomics. However, this is an incorrect assumption. In view of the ongoing advances in mass spectrometry technology, and in particular the ever-increasing mass precision, the elemental isotope definition and its variations should be taken into account. We illustrate the effect of the elemental isotope uncertainty on the theoretical and experimental masses with theoretical calculations and examples.

A hidden Markov-model for gene mapping based on whole-genome next generation sequencing data

The analysis of polygenic, phenotypic characteristics such as quantitative traits or inheritable diseases requires reliable scoring of many genetic markers covering the entire genome. The advent of high-throughput sequencing technologies provides a new way to evaluate large numbers of single nucleotide polymorphisms as genetic markers. Combining the technologies with pooling of segregants, as performed in bulk segregant analysis, should, in principle, allow the simultaneous mapping of multiple genetic loci present throughout the genome. We propose a hidden Markov-model to analyze the marker data obtained by the bulk segregant next generation sequencing. The model includes several states, each associated with a different probability of observing the same/different nucleotide in an offspring as compared to the parent. The transitions between the molecular markers imply transitions between the states of the model. After estimating the transition probabilities and state-related probabilities of nucleotide (dis)similarity, the most probable state for each SNP is selected. The most probable states can then be used to indicate which genomic regions may be likely to contain trait-related genes. The application of the model is illustrated on the data from a study of ethanol tolerance in yeast. Software is written in R. R-functions, R-scripts and documentation are available on www.ibiostat.be/software/bioinformatics.