Introdução: A maior causa de mortes no mundo é devido ao câncer, cerca de 8,2 milhões em 2012 (World Cancer Report, 2014). O câncer de mama é a forma mais comum de câncer entre as mulheres e a segunda neoplasia mais frequente, seguida do câncer de pele não melanoma, representando cerca de 25% de todos os tipos de cânceres diagnosticados. Modelos estatísticos de análise sobrevivência podem ser úteis para a identificação e compreensão de fatores de risco, fatores de prognóstico, bem como na comparação de tratamentos. Métodos: Modelos estatísticos de análise de sobrevivência foram utilizados para evidenciar fatores que afetam os tempos de sobrevida livre da doença e total de um estudo retrospectivo realizado no Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo, Ribeirão Preto, referente a 54 pacientes com câncer de mama localmente avançado com superexpressão do Her-2 que iniciaram a quimioterapia neoadjuvante associada com o medicamento Herceptin® (Trastuzumabe) no período de 2008 a 2012. Utilizaram-se modelos univariados com distribuição Weibull sem e com a presença de fração de cura sob o enfoque frequentista e bayesiano. Utilizou-se modelos assumindo uma estrutura de dependência entre os tempos observados baseados na distribuição exponencial bivariada de Block Basu, na distribuição geométrica bivariada de Arnold e na distribuição geométrica bivariada de Basu-Dhar. Resultados: Resultados da análise univariada sem a presença de covariáveis, o modelo mais adequado às características dos dados foi o modelo Weibull com a presença de fração de cura sob o enfoque bayesiano. Ao incorporar nos modelos as covariáveis, observou-se melhor ajuste dos modelos com fração de cura, que evidenciaram o estágio da doença como um fator que afeta a sobrevida livre da doença e total. Resultados da análise bivariada sem a presença de covariáveis estimam médias de tempo de sobrevida livre da doença para os modelos Block e Basu, Arnold e Basu-Dhar de 108, 140 e 111 meses, respectivamente e de 232, 343, 296 meses para o tempo de sobrevida total. Ao incorporar as covariáveis, os modelos evidenciam que o estágio da doença afeta a sobrevida livre da doença e total. No modelo de Arnold a covariável tipo de cirurgia também se mostrou significativa. Conclusões: Os resultados do presente estudo apresentam alternativas para a análise de sobrevivência com tempos de sobrevida na presença de fração de cura, censuras e várias covariaveis. O modelo de riscos proporcionais de Cox nem sempre se adequa às características do banco de dados estudado, sendo necessária a busca de modelos estatísticos mais adequados que produzam inferências consistentes. / Introduction: The leading worldwide cause of deaths is due to cancer, about 8.2 million in 2012 (World Cancer Report, 2014). Breast cancer is the most common form of cancer among women and the second most common cancer, followed by non-melanoma skin cancer, accounting for about 25% of all diagnosed types of cancers. Statistical analysis of survival models may be useful for the identification and understanding of risk factors, prognostic factors, and the comparison treatments. Methods: Statistical lifetimes models were used to highlight the important factors affecting the disease-free times and the total lifetime about a retrospective study conducted at the Hospital das Clinicas, Faculty of Medicine, University of São Paulo, Ribeirão Preto, referring to 54 patients with locally advanced breast cancer with Her-2 overexpression who started neoadjuvant chemotherapy associated with the drug Herceptin® (Trastuzumab) in the time period ranging from years 2008 to 2012. It was used univariate models assuming Weibull distribution with and without the presence of cure fraction under the frequentist and Bayesian approaches. It was also assumed models assuming a dependence structure between the observed times based on the bivariate Block-Basu exponential distribution, on the bivariate Arnold geometric distribution and on the bivariate Basu-Dhar geometric distribution. Results: From the results of the univariate analysis without the presence of covariates, the most appropriate model for the data was the Weibull model in presence of cure rate under a Bayesian approach. By incorporating the covariates in the models, there was best fit of models with cure fraction, which showed that the stage of the disease was a factor affecting disease-free survival and overall survival. From the bivariate analysis results without the presence of covariates, the estimated means for free survival time of the disease assuming the Block- Basu, Arnold and Basu-Dhar models were respectively given by 108, 140 and 111; for the overall survival times the means were given respectively by, 232, 343, 296 months. In presence of covariates, the models showed that the stage of the disease affects the disease-free survivals and the overall survival times. Assuming the Arnold model, the covariate type of surgery also was significant. Conclusions: The results of this study present alternatives for the analysis of survival times in the presence of cure fraction, censoring and covariates. The Cox proportional hazards model not always is apropriate to the database characteristics studied, which requires the search for more suitable statistical models that produce consistent inferences.

Introdução: A maior causa de mortes no mundo é devido ao câncer, cerca de 8,2 milhões em 2012 (World Cancer Report, 2014). O câncer de mama é a forma mais comum de câncer entre as mulheres e a segunda neoplasia mais frequente, seguida do câncer de pele não melanoma, representando cerca de 25% de todos os tipos de cânceres diagnosticados. Modelos estatísticos de análise sobrevivência podem ser úteis para a identificação e compreensão de fatores de risco, fatores de prognóstico, bem como na comparação de tratamentos. Métodos: Modelos estatísticos de análise de sobrevivência foram utilizados para evidenciar fatores que afetam os tempos de sobrevida livre da doença e total de um estudo retrospectivo realizado no Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo, Ribeirão Preto, referente a 54 pacientes com câncer de mama localmente avançado com superexpressão do Her-2 que iniciaram a quimioterapia neoadjuvante associada com o medicamento Herceptin® (Trastuzumabe) no período de 2008 a 2012. Utilizaram-se modelos univariados com distribuição Weibull sem e com a presença de fração de cura sob o enfoque frequentista e bayesiano. Utilizou-se modelos assumindo uma estrutura de dependência entre os tempos observados baseados na distribuição exponencial bivariada de Block Basu, na distribuição geométrica bivariada de Arnold e na distribuição geométrica bivariada de Basu-Dhar. Resultados: Resultados da análise univariada sem a presença de covariáveis, o modelo mais adequado às características dos dados foi o modelo Weibull com a presença de fração de cura sob o enfoque bayesiano. Ao incorporar nos modelos as covariáveis, observou-se melhor ajuste dos modelos com fração de cura, que evidenciaram o estágio da doença como um fator que afeta a sobrevida livre da doença e total. Resultados da análise bivariada sem a presença de covariáveis estimam médias de tempo de sobrevida livre da doença para os modelos Block e Basu, Arnold e Basu-Dhar de 108, 140 e 111 meses, respectivamente e de 232, 343, 296 meses para o tempo de sobrevida total. Ao incorporar as covariáveis, os modelos evidenciam que o estágio da doença afeta a sobrevida livre da doença e total. No modelo de Arnold a covariável tipo de cirurgia também se mostrou significativa. Conclusões: Os resultados do presente estudo apresentam alternativas para a análise de sobrevivência com tempos de sobrevida na presença de fração de cura, censuras e várias covariaveis. O modelo de riscos proporcionais de Cox nem sempre se adequa às características do banco de dados estudado, sendo necessária a busca de modelos estatísticos mais adequados que produzam inferências consistentes. / Introduction: The leading worldwide cause of deaths is due to cancer, about 8.2 million in 2012 (World Cancer Report, 2014). Breast cancer is the most common form of cancer among women and the second most common cancer, followed by non-melanoma skin cancer, accounting for about 25% of all diagnosed types of cancers. Statistical analysis of survival models may be useful for the identification and understanding of risk factors, prognostic factors, and the comparison treatments. Methods: Statistical lifetimes models were used to highlight the important factors affecting the disease-free times and the total lifetime about a retrospective study conducted at the Hospital das Clinicas, Faculty of Medicine, University of São Paulo, Ribeirão Preto, referring to 54 patients with locally advanced breast cancer with Her-2 overexpression who started neoadjuvant chemotherapy associated with the drug Herceptin® (Trastuzumab) in the time period ranging from years 2008 to 2012. It was used univariate models assuming Weibull distribution with and without the presence of cure fraction under the frequentist and Bayesian approaches. It was also assumed models assuming a dependence structure between the observed times based on the bivariate Block-Basu exponential distribution, on the bivariate Arnold geometric distribution and on the bivariate Basu-Dhar geometric distribution. Results: From the results of the univariate analysis without the presence of covariates, the most appropriate model for the data was the Weibull model in presence of cure rate under a Bayesian approach. By incorporating the covariates in the models, there was best fit of models with cure fraction, which showed that the stage of the disease was a factor affecting disease-free survival and overall survival. From the bivariate analysis results without the presence of covariates, the estimated means for free survival time of the disease assuming the Block- Basu, Arnold and Basu-Dhar models were respectively given by 108, 140 and 111; for the overall survival times the means were given respectively by, 232, 343, 296 months. In presence of covariates, the models showed that the stage of the disease affects the disease-free survivals and the overall survival times. Assuming the Arnold model, the covariate type of surgery also was significant. Conclusions: The results of this study present alternatives for the analysis of survival times in the presence of cure fraction, censoring and covariates. The Cox proportional hazards model not always is apropriate to the database characteristics studied, which requires the search for more suitable statistical models that produce consistent inferences.

The analysis of survival data with a possible cure fraction has attracted much interest in the last two decades. Various models and estimating methods have been proposed for such data and they have been applied in many fields, especially in cancer clinical trials. In the thesis, we consider some new general cure models, which include existing survival models as their special cases. We also consider a nonparametric estimation of cure rate. The estimator is proved consistent and asymptotically normal. We also consider the application of proportional density for cure data and the analysis of length-biased cure data. / Thesis (Ph.D, Mathematics & Statistics) -- Queen's University, 2012-01-26 09:53:08.127

Trois millions de personnes vivent en France avec un antécédent personnel de cancer et ont des difficultés d’accès à l’emprunt et à l’assurance. Depuis 2016, la loi de « modernisation de notre système de santé » a fixé le « droit à l'oubli » (délai au-delà duquel les demandeurs d’assurance ayant eu un antécédent de cancer n’auront plus à le déclarer) à 10 ans après la fin des traitements. D’un point de vue statistique, on peut considérer ce délai comme le délai au-delà duquel la surmortalité liée au cancer (taux de mortalité en excès) s’annule durablement, ce qui se traduit sur les courbes de survie nette par un plateau correspondant à la proportion de patients guéris. La vérification de l’hypothèse de guérison repose sur deux critères : un taux de mortalité en excès négligeable et la confirmation graphique de l’existence d’un plateau. Une nouvelle définition du délai de guérison a été proposée pour ce travail comme le temps à partir duquel la probabilité d’appartenir au groupe des guéris atteint 95%.Le premier objectif de cette thèse était de fournir des estimations du délai de guérison à partir des données des registres de cancer du réseau FRANCIM pour chaque localisation de cancer selon le sexe et l’âge. Le délai de guérison est inférieur à 12 ans pour la majorité des localisations vérifiant l’hypothèse de guérison. Il est notamment inférieur ou égal à 5 ans, voire nul pour certaines classes d’âge, pour le mélanome de la peau, le cancer du testicule et de la thyroïde. Les critères pour la vérification de la guérison sont subjectifs et le délai de guérison ne repose pas sur une estimation directe par les modèles de guérison préexistants. Un nouveau modèle de guérison a été développé, incluant le délai de guérison comme paramètre à estimer afin de répondre objectivement à la question de l’existence d’une guérison statistique et de permettre une estimation directe du délai de guérison.Le second objectif de la thèse était de comparer, dans des situations contrôlées pour lesquelles le taux de mortalité en excès devenait nul, les performances de ce nouveau modèle à celles de deux autres modèles de guérison. La survie nette et la proportion de guéris estimées par les modèles ont été comparées aux valeurs théoriques utilisées pour simuler les données. Le nouveau modèle permet, avec des conditions strictes d’application, d’estimer directement le délai de guérison avec des performances aussi satisfaisantes que celles des autres modèles. / Three million people are living in France with a personal past of cancer and undergo difficulties in accessing loans and insurance. Since 2016, the French law « modernisation de notre système de santé » set the "right to be forgotten" (time beyond which insurance applicants with a past of cancer will no longer have to declare it) at 10 years after the end of treatment. From a statistical point of view, this delay can be considered as the time from which mortality due to cancer (excess mortality) disappears. After this time, the net survival curves reach a plateau corresponding to the proportion of cured patients. The verification of this hypothesis is based on two criteria: a negligible excess mortality rate and a graphic confirmation of the existence of a plateau. We proposed a new definition of the time-to-cure as the time from which the probability of belonging to the cured group reaches 95%.The first aim of this thesis was to estimate time-to-cure for each cancer site by sex and age using population-based data from the FRANCIM registries network. Time-to-cure was lower than 12 years in most sites complying with the cure hypothesis. It was less than 5 years, or even null in some age groups, for skin melanoma, testicular and thyroid. Criteria to verify the cure hypothesis are subjective and time-to-cure is not directly estimated in pre-existing cure models. A new model has been developed including time-to-cure as a parameter to address the question of statistical cure and to allow direct estimation of time-to-cure.The second objective of this thesis was to compare, in controlled situations in which the excess mortality rate became null, the performances of this new model with that of two other cure models. Estimated net survival and cure fraction have been compared to the theoretical values used to simulate the data. Direct estimation of time-to-cure is possible under strict conditions.

Survival analytic methods help to analyze lifetime data arising from medical and reliability experiments. The popular proportional hazards model, proposed by Cox (1972), is widely used in survival analysis to study the effect of risk factors on lifetimes. An important assumption in regression type analysis is that all relative risk factors should be included in the model. However, not all relative risk factors are observed due to measurement difficulty, inaccessibility, cost considerations, and so on. These unobservable risk factors can be modelled by the so-called frailty model, originally introduced by Vaupel et al. (1979). Furthermore, the frailty model is also applicable to clustered data. Cluster data possesses the feature that observations within the same cluster share similar conditions and environment, which are sometimes difficult to observe. For example, patients from the same family share similar genetics, and patients treated in the same hospital share the same group of profes- sionals and same environmental conditions. These factors are indeed hard to quantify or measure. In addition, this type of similarity introduces correlation among subjects within clusters. In this thesis, a semi-parametric frailty model is proposed to address aforementioned issues. The baseline hazards function is approximated by a piecewise constant function and the frailty distribution is assumed to be a Birnbaum-Saunders distribution. Due to the advancement in modern medical sciences, many diseases are curable, which in turn leads to the need of incorporating cure proportion in the survival model. The frailty model discussed here is further extended to a mixture cure rate frailty model by integrating the frailty model into the mixture cure rate model proposed originally by Boag (1949) and Berkson and Gage (1952). By linking the covariates to the cure proportion through logistic/logit link function and linking observable covariates and unobservable covariates to the lifetime of the uncured population through the frailty model, we obtain a flexible model to study the effect of risk factors on lifetimes. The mixture cure frailty model can be reduced to a mixture cure model if the effect of frailty term is negligible (i.e., the variance of the frailty distribution is close to 0). On the other hand, it also reduces to the usual frailty model if the cure proportion is 0. Therefore, we can use a likelihood ratio test to test whether the reduced model is adequate to model the given data. We assume the baseline hazard to be that of Weibull distribution since Weibull distribution possesses increasing, constant or decreasing hazard rate, and the frailty distribution to be Birnbaum-Saunders distribution. D ́ıaz-Garc ́ıa and Leiva-Sa ́nchez (2005) proposed a new family of life distributions, called generalized Birnbaum-Saunders distribution, including Birnbaum-Saunders distribution as a special case. It allows for various degrees of kurtosis and skewness, and also permits unimodality as well as bimodality. Therefore, integration of a generalized Birnbaum-Saunders distribution as the frailty distribution in the mixture cure frailty model results in a very flexible model. For this general model, parameter estimation is carried out using a marginal likelihood approach. One of the difficulties in the parameter estimation is that the likelihood function is intractable. The current technology in computation enables us to develop a numerical method through Monte Carlo simulation, and in this approach, the likelihood function is approximated by the Monte Carlo method and the maximum likelihood estimates and standard errors of the model parameters are then obtained numerically by maximizing this approximate likelihood function. An EM algorithm is also developed for the Birnbaum-Saunders mixture cure frailty model. The performance of this estimation method is then assessed by simulation studies for each proposed model. Model discriminations is also performed between the Birnbaum-Saunders frailty model and the generalized Birnbaum-Saunders mixture cure frailty model. Some illustrative real life examples are presented to illustrate the models and inferential methods developed here. / Thesis / Doctor of Science (PhD)