京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第21681号 / 医博第4487号 / 新制||医||1036(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 松田 文彦, 教授 清水 章, 教授 濵﨑 洋子 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Use of progression-free survival in the evaluation of clinical interventions is hampered by a variety of issues, including censoring patterns not addressed in the usual methods for survival analysis. Progression can be right-censored before survival or interval-censored between inspection times. Current practice calls for imputing events to their time of detection. Such an approach is prone to bias, underestimates standard errors and makes inefficient use of the data at hand. Moreover a composite outcome prevents inference about the actual treatment effect on the risk of progression. This thesis develops semiparametric and sieve maximum likelihood estimators to more formally analyze progression-related endpoints. For the special case where death rarely precedes progression, a Cox-Aalen model is proposed for regression analysis of time-to-progression under intermittent inspection. The general setting considering both progression and survival is examined with a Markov Cox-type illness-death model under various censoring schemes. All of the resulting estimators globally converge to the truth slower than the parametric rate, but their finite-dimensional components are asymptotically efficient. Numerical studies suggest that the new methods perform better than their imputation-based alternatives under moderate to large samples having higher rates of censoring.

L'analyse de survie constitue bien souvent l'objectif principal des études cliniques en cancérologie. Les données de survie découlent d'un événement subi par les sujets de l'étude, événement qui correspond par exemple au décès pour la survie globale et à la progression tumorale pour la survie sans progression. Les méthodes non-paramétriques de Kaplan-Meier et semi-paramétriques de Cox représentent les modèles standards les plus utilisés pour modéliser ces données de survie, mais ne s'appliquent que dans le cas d'un seul événement temporel. La survie globale est considérée comme le critère clinique optimal pour juger de l'efficacité d'un traitement. La survie sans progression est un critère intermédiaire, qui représente un critère potentiel de substitution pour la survie globale. Depuis plusieurs années, un intérêt croissant s'est porté sur la validation statistique de critères intermédiaires. Cette validation passe par la mesure de la corrélation entre le critère clinique principal et le critère intermédiaire. Ainsi, une modélisation bivariée apparait intéressante afin de décrire la structure de dépendance entre les survies sans progression et globale. L'objectif de cette thèse concerne la modélisation de la structure d'association entre les survies sans progression et globale ainsi que la quantification de cette association via des mesures de dépendance. Pour cela, nous étudions en premier lieu les extensions du modèle de Cox qui peuvent traiter la dépendance statistique entre les données. Nous proposons ensuite une nouvelle modélisation paramétrique de la survie globale basée sur une distribution conditionnelle et sur les survies sans progression et post-progression. De plus, nous examinons différents modèles paramétriques de survie bivariée en termes de mesures de corrélation. Ces modèles sont fondés sur deux approches : les distributions marginales et l'indépendance conditionnelle. Enfin, nous appliquons et comparons les modèles étudiés en utilisant les données d'un essai clinique randomisé de phase III, impliquant des patients atteints de cancer du poumon non à petites cellules localement avancé. / Analysis of survival often represents the main aim in cancer clinical studies. Survival data arise from an event experienced by the study subjects. This event corresponds for example to the death for overall survival and to tumor progression for progression-free survival. The Kaplan-Meier nonparametric estimator and the Cox semiparametric model are the most used standard methods for modeling survival data, although they are applied only in the case of unique temporal event. Overall survival is the optimal clinical endpoint for assessing the efficiency of treatment. Progression-free survival is an intermediate endpoint considered as a potential surrogate of overall survival. For the past few years, we observed an increasing focus on statistical validation of intermediate endpoints and this through measurement of the correlation between the principal clinical endpoint and the intermediate one. Thus, bivariate modeling could be of interest for describing the dependence structure between progression-free survival and overall survival. The aim of this thesis is the modeling of the structure of association between progression-free survival and overall survival as well as the quantification of this association using dependence measures. For this, we study at first extensions of Cox model able to address the topic concerning the statistical dependence between data. Next, we propose a new parametric modeling of overall survival based on two survival times : progression-free survival and post-progression survival, assumed to be linked by a conditional distribution. Moreover, we examine different parametric models for bivariate survival data concerning correlation measurement. These models are based on the marginal distributions and the conditional independence. Finally, we apply and compare these models using data from a phase III randomized clinical trial, involving patients with locally advanced non-small cell lung cancer.

Le mésothéliome pleural malin (MPM) est un cancer rare et de mauvais pronostic. Le diagnostic de MPM restant difficile à ce jour, nous avons tout d'abord évalué l'intérêt de l'analyse du transcriptome, à partir des cellules présentes dans le liquide pleural, à des fins diagnostiques. Si cette approche parait faisable, ses contraintes techniques la rendent incompatibles avec une utilisation en routine. Notre deuxième axe de travail s'est inscrit dans le domaine du traitement local du MPM, avec l'évaluation de l'administration intrapleurale (IP) de deux molécules de chimiothérapie : le pemetrexed et le lipoplatine. Les résultats sur modèle animal montrent que les profils pharmacocinétiques de ces deux molécules sont significativement différents entre une administration intraveineuse et une administration IP. Dans l'optique d'identifier de nouvelles cibles thérapeutiques, nous avons évalué la voie de l'adrénomédulline (AM) dans le MPM. L'AM et ses récepteurs sont conjointement exprimés dans les biopsies pleurales de MPM. In vitro, l'AM augmente les capacités de prolifération, de migration et d'invasion des cellules de MPM, par l'intermédiaire de la voie des MAPK. L'inhibition de l'AM ou de ses récepteurs apparait comme une stratégie thérapeutique prometteuse dans le MPM, de part ses effets directs sur les cellules néoplasiques, mais aussi ses effets indirects, via une inhibition de l'angiogénèse et de la lymphangiogénèse tumorale (in vivo). Enfin, nous avons évalué de nouveaux critères de jugement pour les essais cliniques de phase II. Nous avons montré que le taux de survie sans progression à 9 semaines est le critère le plus performant pour prédire la survie globale. / Malignant pleural mesothelioma (MPM) is a rare cancer with poor prognosis. As MPM diagnosis remains difficult today, we first assessed the potential value of transcriptome analysis, from cells in pleural fluid, with diagnostic purpose. If this approach looks feasible, its technical constraints make it incompatible with routine practice. The second axis of our work concerned MPM local treatment, with the assessment of intrapleural (IP) administration of two chemotherapy drugs: pemetrexed and lipoplatin. Results obtained in an animal model show that the pharmacokinetic profiles of these two drugs are significantly different between intravenous and IP administration. With the goal of identifying new therapeutic targets, we explored the adrenomedullin (AM) pathway in MPM. AM and its receptors are jointly expressed in pleural biopsies from MPM. Moreover we demonstrated in vitro that AM increases the proliferation, migration and invasion abilities of MPM cells, through the MAPK signaling pathway. Inhibition of AM or its receptors appears as a promising therapeutic strategy, because of its direct effects on malignant cells, but also its indirect effects, via tumor angiogenesis and lymphangiogenesis inhibition (in vivo). Finally, we assessed new endpoints for phase II clinical studies. We showed that the progression-free survival rate at 9 weeks is the most performant criterion to predict overall survival.

Background: Glioblastoma multiforme (GBM) is the most common and aggressive brain tumor with poor clinical outcome. Identification and development of new markers could be beneficial for the diagnosis and prognosis of GBM patients. Deregulation of microRNAs (miRNAs or miRs) is involved in GBM. Therefore, we attempted to identify and develop specific miRNAs as prognostic and predictive markers for GBM patient survival.Methods: Expression profiles of miRNAs and genes and the corresponding clinical information of 480 GBM samples from The Cancer Genome Atlas (TCGA) dataset were downloaded and interested miRNAs were identified. Patients' overall survival (OS) and progression-free survival (PFS) associated with interested miRNAs and miRNA-interactions were performed by Kaplan-Meier survival analysis. The impacts of miRNA expressions and miRNA-interactions on survival were evaluated by Cox proportional hazard regression model. Biological processes and network of putative and validated targets of miRNAs were analyzed by bioinformatics.Results: In this study, 6 interested miRNAs were identified. Survival analysis showed that high levels of miR-326/miR-130a and low levels of miR-323/miR-329/miR-155/miR-210 were significantly associated with long OS of GBM patients, and also showed that high miR-326/miR-130a and low miR-155/miR-210 were related with extended PFS. Moreover, miRNA-323 and miRNA-329 were found to be increased in patients with no-recurrence or long time to progression (TTP). More notably, our analysis revealed miRNA-interactions were more specific and accurate to discriminate and predict OS and PFS. This interaction stratified OS and PFS related with different miRNA levels more detailed, and could obtain longer span of mean survival in comparison to that of one single miRNA. Moreover, miR-326, miR-130a, miR-155, miR-210 and 4 miRNA-interactions were confirmed for the first time as independent predictors for survival by Cox regression model together with clinicopathological factors: Age, Gender and Recurrence. Plus, the availability and rationality of the miRNA-interaction as predictors for survival were further supported by analysis of network, biological processes, KEGG pathway and correlation analysis with gene markers.Conclusions: Our results demonstrates that miR-326, miR-130a, miR-155, miR-210 and the 4 miRNA-interactions could serve as prognostic and predictive markers for survival of GBM patients, suggesting a potential application in improvement of prognostic tools and treatments.

Background: Glioblastoma multiforme (GBM) is the most common and aggressive brain tumor with poor clinical outcome. Identification and development of new markers could be beneficial for the diagnosis and prognosis of GBM patients. Deregulation of microRNAs (miRNAs or miRs) is involved in GBM. Therefore, we attempted to identify and develop specific miRNAs as prognostic and predictive markers for GBM patient survival.Methods: Expression profiles of miRNAs and genes and the corresponding clinical information of 480 GBM samples from The Cancer Genome Atlas (TCGA) dataset were downloaded and interested miRNAs were identified. Patients' overall survival (OS) and progression-free survival (PFS) associated with interested miRNAs and miRNA-interactions were performed by Kaplan-Meier survival analysis. The impacts of miRNA expressions and miRNA-interactions on survival were evaluated by Cox proportional hazard regression model. Biological processes and network of putative and validated targets of miRNAs were analyzed by bioinformatics.Results: In this study, 6 interested miRNAs were identified. Survival analysis showed that high levels of miR-326/miR-130a and low levels of miR-323/miR-329/miR-155/miR-210 were significantly associated with long OS of GBM patients, and also showed that high miR-326/miR-130a and low miR-155/miR-210 were related with extended PFS. Moreover, miRNA-323 and miRNA-329 were found to be increased in patients with no-recurrence or long time to progression (TTP). More notably, our analysis revealed miRNA-interactions were more specific and accurate to discriminate and predict OS and PFS. This interaction stratified OS and PFS related with different miRNA levels more detailed, and could obtain longer span of mean survival in comparison to that of one single miRNA. Moreover, miR-326, miR-130a, miR-155, miR-210 and 4 miRNA-interactions were confirmed for the first time as independent predictors for survival by Cox regression model together with clinicopathological factors: Age, Gender and Recurrence. Plus, the availability and rationality of the miRNA-interaction as predictors for survival were further supported by analysis of network, biological processes, KEGG pathway and correlation analysis with gene markers.Conclusions: Our results demonstrates that miR-326, miR-130a, miR-155, miR-210 and the 4 miRNA-interactions could serve as prognostic and predictive markers for survival of GBM patients, suggesting a potential application in improvement of prognostic tools and treatments.

In general population 1 of 72 women develop ovarian cancer and to 1 of 95 women this disease is lethal. A great number of clinical trials have shown that the course of the disease is not dependent only on the classical prognostic indicators such as histological tumor type, tumor differentiation, stage of the disease or treatment modalities. More than two decades ago the first publications on heredity factors indicated similarity among the patients diagnosed ovarian malignancies and their first degree relatives. The first genetic autosomal dominant inheritance was determined in the high-risk cancer tumor suppressor BRCA1/2 genes. In spite of the abundant number of trials studying the BRCA1/2 genes role in breast and ovarian carcinogenesis still it is not sufficiently clear the influence of these genes for the disease prognosis. The aim of our conducted trial was to determine the BRCA1/2 genes prognostic significance for progression-free and overall survival in the event of advanced ovarian cancer. In case of advanced ovarian cancer the BRCA1/2 mutation frequency was 51,4 %. Among all determined BRCA1/2 gene mutations BRCA1 4035delA or founder mutation was most frequent. It amounted to 63.6%. Non-optimal cytoreduction (p<0,0001 ) , patients’ older age (p=0,005) and absence of BRCA1/2 mutations (p=0,049) are closely connected with a shorter PFS and OS. Only non-optimal cytoreduction was related to a shorter OS (p=0,010). / Bendrojoje populiacijoje 1 iš 72 moterų suserga kiaušidžių vėžiu ir 1 iš 95 moterų miršta nuo šios ligos. Tyrimų duomenys rodo, kad ligos eiga nėra priklausoma vien tik nuo klasikinių prognozinių rodiklių, tokių kaip histologinis naviko tipas, naviko diferenciacija, ligos stadija, taikytas gydymas.Prognozinių veiksnių paieška krypstą link genetinių veiksnių galinčių įtakoti ligos eigą. Literatūros duomenys apie klinikinę BRCA1/2 genų reikšmę yra kontroversiški – nuo visiškai bereikšmio iki ženkliai teigiamo poveikio ligos eigai prognoziniu požiūriu.. Mūsų tyrėjų grupės atlikto tyrimo tikslas buvo nustatyti BRCA1/2 genų mutacijų dažnį ir jų įvairovę tarp pacienčių, sergančių išplitusiu kiaušidžių vėžiu, ir įvertinti šių mutacijų įtaką berecidyviam ir bendrajam išgyvenamumui. Mes nustatėme , kad tarp pacienčių sergančių išplitusių epiteliniu kiaušidžių vėžiu buvo net 51,4 proc. BRCA 1/2 mutacijų genuose turinčių pacienčių. 98,2 proc. šių pacienčių sirgo serozine papiline adenokarcinoma. Šios histologinės formos kiaušidžių vėžio buvo ženkliai daugiau mutuotų BRCA1/2 genų pacienčių grupėje nei tarp pacienčių be mutacijų (p-0,029). Tyrimo metu nustatėme dažniausiai sutinkamą arba bendro protėvio BRCA 1 4035 delA mutaciją bei taip kad statistiškai reikšmingos įtakos sergančiųjų išplitusiu kiaušidžių vėžiu berecidyviam išgyvenamumui turi pacienčių amžius (p=0,005), BRCA1/2 genų mutacijos(p=0,049) bei operacijos apimtis (p<0,0001), o bendrajam išgyvenamumui – tik operacijos... [toliau žr. visą tekstą]

Bendrojoje populiacijoje 1 iš 72 moterų suserga kiaušidžių vėžiu ir 1 iš 95 moterų miršta nuo šios ligos. Tyrimų duomenys rodo, kad ligos eiga nėra priklausoma vien tik nuo klasikinių prognozinių rodiklių, tokių kaip histologinis naviko tipas, naviko diferenciacija, ligos stadija, taikytas gydymas.Prognozinių veiksnių paieška krypstą link genetinių veiksnių galinčių įtakoti ligos eigą. Literatūros duomenys apie klinikinę BRCA1/2 genų reikšmę yra kontroversiški – nuo visiškai bereikšmio iki ženkliai teigiamo poveikio ligos eigai prognoziniu požiūriu.. Mūsų tyrėjų grupės atlikto tyrimo tikslas buvo nustatyti BRCA1/2 genų mutacijų dažnį ir jų įvairovę tarp pacienčių, sergančių išplitusiu kiaušidžių vėžiu, ir įvertinti šių mutacijų įtaką berecidyviam ir bendrajam išgyvenamumui. Mes nustatėme , kad tarp pacienčių sergančių išplitusių epiteliniu kiaušidžių vėžiu buvo net 51,4 proc. BRCA 1/2 mutacijų genuose turinčių pacienčių. 98,2 proc. šių pacienčių sirgo serozine papiline adenokarcinoma. Šios histologinės formos kiaušidžių vėžio buvo ženkliai daugiau mutuotų BRCA1/2 genų pacienčių grupėje nei tarp pacienčių be mutacijų (p-0,029). Tyrimo metu nustatėme dažniausiai sutinkamą arba bendro protėvio BRCA 1 4035 delA mutaciją bei taip kad statistiškai reikšmingos įtakos sergančiųjų išplitusiu kiaušidžių vėžiu berecidyviam išgyvenamumui turi pacienčių amžius (p=0,005), BRCA1/2 genų mutacijos(p=0,049) bei operacijos apimtis (p<0,0001), o bendrajam išgyvenamumui – tik operacijos... [toliau žr. visą tekstą] / In general population 1 of 72 women develop ovarian cancer and to 1 of 95 women this disease is lethal. A great number of clinical trials have shown that the course of the disease is not dependent only on the classical prognostic indicators such as histological tumor type, tumor differentiation, stage of the disease or treatment modalities. More than two decades ago the first publications on heredity factors indicated similarity among the patients diagnosed ovarian malignancies and their first degree relatives. The first genetic autosomal dominant inheritance was determined in the high-risk cancer tumor suppressor BRCA1/2 genes. In spite of the abundant number of trials studying the BRCA1/2 genes role in breast and ovarian carcinogenesis still it is not sufficiently clear the influence of these genes for the disease prognosis. The aim of our conducted trial was to determine the BRCA1/2 genes prognostic significance for progression-free and overall survival in the event of advanced ovarian cancer. In case of advanced ovarian cancer the BRCA1/2 mutation frequency was 51,4 %. Among all determined BRCA1/2 gene mutations BRCA1 4035delA or founder mutation was most frequent. It amounted to 63.6%. Non-optimal cytoreduction (p<0,0001 ) , patients’ older age (p=0,005) and absence of BRCA1/2 mutations (p=0,049) are closely connected with a shorter PFS and OS. Only non-optimal cytoreduction was related to a shorter OS (p=0,010).

No description available.

De nos jours, il est de plus en plus fréquent de recourir à des mesures intermédiaires d’efficacité telles que la survie sans progression (SSP) et le temps avant la progression de la tumeur (TPT) afin d’estimer l’efficacité d’un nouvel agent anticancéreux. Cependant, l’absence de mesures finales comme la survie globale (SG) complexifie la prise de décision par rapport au remboursement des nouveaux traitements anticancéreux. Ainsi, cette thèse se concentre sur différents aspects de l’intégration des mesures intermédiaires d’efficacité dans les évaluations économiques en oncologie, spécifiquement en cancer du sein métastatique. Une première étude a évalué la relation entre la SSP/TPT et la SG dans le contexte du cancer du sein métastatique. Une revue systématique de la littérature a identifié les études cliniques randomisées portant sur l’efficacité d’un traitement anticancéreux chez les femmes atteintes d’un cancer du sein métastatique et rapportant des données de SSP/TPT et de SG. Les résultats de cette étude ont démontré qu’il existe une relation statistiquement significative, d’une part, entre la SSP/TPT médiane et la SG médiane (r = 0.428; p ≤ 0,01), et d’autre part, entre l’effet de traitement sur la SSP/TPT et l’effet de traitement sur la SG (r = 0.427; p ≤ 0,01). Selon les résultats obtenus, la SSP/TPT pourrait être considérée comme un substitut valide de la SG, justifiant ainsi son utilisation dans les évaluations économiques en cancer du sein métastatique. Une deuxième étude a évalué l’utilisation des mesures intermédiaires de survie dans les évaluations économiques en cancer avancé et identifié les méthodes utilisées pour intégrer ces mesures en l’absence de données de SG. Une revue systématique de la littérature a été réalisée pour recenser les évaluations économiques de type coût-efficacité et coût-utilité ayant intégré des mesures intermédiaires de survie. Cette étude a démontré l’ampleur de l’utilisation des mesures intermédiaires de survie dans les évaluations économiques en cancer avancé. Par ailleurs, plusieurs approches ont été identifiées pour pallier l’absence de données de SG, notamment l’utilisation d’un risque de décès post-progression équivalent pour les groupes à l’étude, le recours à des comparaisons indirectes basées sur de nombreuses hypothèses, l’utilisation d’une mesure intermédiaire comme proxy de la SG, le recours à l’opinion d’experts et l’utilisation de données associées à un traitement différent ou une ligne de traitement différente. Enfin, une troisième étude s’est penchée sur le développement d’un modèle pharmacoéconomique générique canadien intégrant les mesures intermédiaires de survie en cancer du sein métastatique. Ce modèle de Markov inclut des paramètres spécifiques aux traitements sous évaluations (coût de traitement, données de survie et incidence des effets indésirables) de même que des paramètres globaux qui ne dépendent pas des traitements évalués (caractéristiques des patientes, valeurs d’utilité associées aux états de santé du modèle, pertes d’utilité et coûts des effets indésirables, coûts d’administration des traitements, coûts de suivi médical et coûts des soins prodigués en fin de vie). Le modèle a été validé en évaluant sa capacité à répliquer des résultats d’études existantes. Ce modèle permet d’uniformiser l’évaluation économique des nouveaux traitements en cancer du sein métastatique et pourrait par conséquent, devenir un outil d’aide à la décision de référence pour les organismes responsables du remboursement des médicaments au Canada. Bref, les résultats de ces trois études répondent à une problématique importante dans l’évaluation économique des traitements en oncologie et pourront contribuer à faciliter la prise de décision en santé. / Nowadays, intermediate endpoints such as progression-free survival (PFS) and time to progression (TTP) are frequently used in clinical trials of advanced cancer. However, use of such endpoints instead of overall survival (OS) poses a significant challenge in the economic evaluation of anticancer drugs. This thesis focuses on different aspects of the integration of intermediate endpoints in the economic evaluation of anticancer drugs, especially in the context of metastatic breast cancer. A first study assessed the relationship between PFS/TPT and OS in metastatic breast cancer using a trial-based approach. A systematic review of the literature was performed to identify randomized clinical trials of metastatic breast cancer therapy reporting both PFS/TTP and OS data. Results of this study indicated a statistically significant relationship between the median PFS/TTP and the median OS (r = 0.428; p < 0.01), and between the treatment effect on PFS/TTP and the treatment effect on OS (r = 0.427; p < 0.01). Findings of this study suggest that PFS/TTP may be considered as a potential surrogate for OS, thus justifying its use in cost-effectiveness or cost-utility analyses of metastatic breast cancer therapy. A second study evaluated the use of intermediate endpoints in the economic evaluation of new treatments for advanced cancer and the methodological approaches adopted when OS data are unavailable or of limited use. A systematic review of the literature was conducted to identify cost-effectiveness and cost-utility analyses using an intermediate endpoint as an outcome measure. This study showed that intermediate endpoints are widely used in the economic evaluation of new treatments for advanced cancer. Several approaches are used in the absence of OS data such as assuming an equal risk of death for all treatment groups, using indirect comparison based on numerous assumptions, using of a proxy for OS, using unpublished external information (consultation with clinical experts), and using published external information from different treatment settings. Finally, a third study aimed to develop a global economic model to assess the cost-effectiveness of new treatments for metastatic breast cancer in Canada. This Markov model, which integrates intermediate endpoints, includes parameters specific to the treatments under evaluation (drug treatment, survival outcomes, and incidence of treatment-related adverse events (AEs)), as well as global parameters that are consistent regardless of the treatment under evaluation (patient characteristics, health states utilities, disutilities and costs associated with treatment-related AEs, as well as costs associated with drug administration, medical follow-up, and end-of-life care). The model was validated by assessing its ability to replicate results of existing studies. This model standardizes the economic evaluation of new therapies for metastatic breast cancer, and could thus be used as a benchmark by drug reimbursement authorities in Canada. In summary, the results of these three studies address an important challenge encountered in the economic evaluation of anticancer drugs, and therefore, can be very valuable for decision-making purposes.