Background: Black South African women are known to have a high usage rate of injectable contraceptives. Breast cancer is the second leading cancer after malignant cervical neoplasms in black South African women. There is evidence that sex hormones are associated with an increased risk of developing breast cancer. In the Western Cape, investigators suggested that injectable contraceptives, more specifically DMPA, may increase breast cancer risk. In another study conducted in the same province, a weak association between breast cancer and women taking combined oestrogen/progesterone oral contraceptives was found, though no risk associated with injectable progestogen contraceptives (DMPA) was confirmed. Study Objective: This study aimed to determine whether there is an association between hormonal contraceptive use and an increased risk of cancer of the breast. Methods: Data was obtained from an ongoing case control study set up by MRC/Wits/NHLS Cancer Epidemiology Research Group (CERG) in 1995 to investigate risk factors associated with cancer among the black population in Johannesburg. Data was processed using STATA, version8 and analysed using univariate, bivariate and multivariate unmatched logistic regression models. Results: There was evidence that an overall use of oral contraceptives increases the risk of breast cancer; cases (n= 221), controls :( n= 153), OR=2.01 (95% CI:1.45, 2.80), p<0.0001. There was evidence of an association between use of injectable contraception and the risk of breast cancer; cases (n=244), controls (n=202), OR=1.51(CI: 1.14, 2.01),p=0.004 Surprisingly, no other use characteristic of either hormonal contraceptive method was statistically significantly associated with the risk of breast cancer in our dataset. The combined use of both oral and injectable contraception was associated with an increased risk of breast cancer, OR=1.68(1.21, 2.33), p =0.002. There was a strong effect modification (interaction) between oral contraceptive use and injectable progesterone associated with the risk of breast cancer, (p=0.008). Conclusion: After adjusting for all potential risk and confounding factors, as collected in the dataset, there was evidence of an association between combined oral contraceptive use and breast cancer. An association between cancer of the breast and overall use of injectable progesterone use was also established. There was evidence of association between the use of both hormonal contraceptive methods and an increased risk of breast cancer. However, whether these findings reflect the reality in terms of causal relationship or are the result of bias must be ascertained.

En recherche médicale, les cohortes permettent de mieux comprendre l'évolution d'une pathologie et d'améliorer la prise en charge des patients. La mise en évidence de liens de causalité entre certains facteurs de risque et l'évolution de l'état de santé des patients est possible grâce à des études étiologiques. L'analyse de cohortes permet aussi d'identifier des marqueurs pronostiques de l'évolution d'un état de santé. Cependant, les facteurs de confusion constituent souvent une source de biais importante dans l'interprétation des résultats des études étiologiques ou pronostiques. Dans ce manuscrit, nous présentons deux travaux de recherche en Biostatistique dans la thématique des scores de propension. Dans le premier travail, nous comparons les performances de différents modèles permettant d'évaluer la causalité d'une exposition sur l'incidence d'un événement en présence de données censurées à droite. Dans le second travail, nous proposons un estimateur de courbes ROC dépendantes du temps standardisées et pondérées permettant d'estimer la capacité prédictive d'un marqueur en prenant en compte les facteurs de confusion potentiels.En cohérence avec l'objectif de fournir des outils statistiques adaptés, nous présentons également dans ce manuscrit une application nommée Plug-Stat®. En lien direct avec la base de données, elle permet de réaliser des analyses statistiques adaptées à la pathologie afin de faciliter la recherche épidémiologique et de mieux valoriser les données de cohortes observationnelles. / In medical research, cohorts help to better understandthe evolution of a pathology and improve the care ofpatients. Causal associations between risk factors andoutcomes are regularly studied through etiological studies. Cohorts analysis also allow the identification of new markers for the prediction of the patient evolution.However, confounding factors are often source of bias in the interpretation of the results of etiologic or prognostic studies.In this manuscript, we presented two research works in Biostatistics, the common topic being propensity scores.In the first work, we compared the performances of different models allowing to evaluate the causality of an exposure on an outcome in the presence of rightc ensored data. In the second work, we proposed anestimator of standardized and weighted time-dependentROC curves. This estimator provides a measure of theprognostic capacities of a marker by taking into accountthe possible confounding factors. Consistent with our objective to provide adapted statistical tools, we also present in this manuscript an application, so-calledPlug-Stat®. Directly linked with the database, it allows toperform statistical analyses adapted to the pathology in order to facilitate epidemiological studies and improve the valorization of data from observational cohorts.

Dementia is a clinical syndrome caused by neurodegeneration or cerebrovascular injury. Patients with dementia suffer from deterioration in memory, thinking, behavior and the ability to perform everyday activities. Since there are no cures or disease-modifying therapies for dementia, there is much interest in identifying modifiable risk factors that may help prevent or slow the progression of cognitive decline. Medications are a common focus of this type of research. Importantly, according to a report from the Centers for Disease Control and Prevention (CDC), 19.1% of the population aged 60 and over report taking antidepressants during 2011-2014, and this number tends to increase. However, antidepressant use among the elderly may be concerning because of the potentially harmful effects on cognition. To assess the impacts of antidepressants on the risk of dementia, we conducted three consecutive projects. In the first project, a retrospective cohort study using Marginal Structural Cox Proportional Hazards regression model with Inverse Probability Weighting (IPW) was conducted to evaluate the average causal effects of different classes of antidepressant on the risk of dementia. Potential causal effects of selective serotonin reuptake inhibitors (SSRIs), serotonin and norepinephrine reuptake inhibitors (SNRIs), atypical anti-depressants (AAs) and tri-cyclic antidepressants (TCAs) on the risk of dementia were observed at the 0.05 significance level. Multiple sensitivity analyses supported these findings. Unmeasured confounding is a threat to the validity of causal inference methods. In evaluating the effects of antidepressants, it is important to consider how common comorbidities of depression, such as sleep disorders, may affect both the exposure to anti-depressants and the onset of cognitive impairment. In this dissertation, sleep apnea and rapid-eye-movement behavior disorder (RBD) were unmeasured and thus uncontrolled confounders for the association between antidepressant use and the risk of dementia. In the second project, a bias factor formula for two binary unmeasured confounders was derived in order to account for these variables. Monte Carlo analysis was implemented to estimate the distribution of the bias factor for each class of antidepressant. The effects of antidepressants on the risk of dementia adjusted for both measured and unmeasured confounders were estimated. Sleep apnea and RBD attenuated the effect estimates for SSRI, SNRI and AA on the risk of dementia. In the third project, to account for potential time-varying confounding and observed time-varying treatment, a multi-state Markov chain with three transient states (normal cognition, mild cognitive impairment (MCI), and impaired but not MCI) and two absorbing states (dementia and death) was performed to estimate the probabilities of moving between finite and mutually exclusive cognitive state. This analysis also allowed participants to recover from mild impairments (i.e., mild cognitive impairment, impaired but not MCI) to normal cognition, and accounted for the competing risk of death prior to dementia. These findings supported the results of the main analysis in the first project.

Draper and Guttman (1997) show that for basic 2^{k-p}designs, p >= 0, k-p replicates of blocks designs of size two are needed to estimate all the usual (estimable) effects. In Chapter 1, we provide an algebraic formal proof for the two-level blocks designs results and present results applicable to the general case; that is, for the case of s^{k} factorial (p=0) or s^{k-p}ractional factorial (p>0) designs in s^{b} blocks, where 0<b<k-p, at least (k-p)/(k-p-b) replicates are needed to clear up all possible effects. Through the theoretical development presented in this work, it can provide a clearer view on why those results would hold. We will also discuss the estimation equations given in Draper and Guttman (1997). In Chapter 2, we present a methodology for analyzing the variability of the radiated emission testings of electronic, elecommunication and information technology equipment based on a modified analysis of variance (ANOVA), with polynomial regression analysis. In our study, three electronic products; modem, monitor and notebook bought from the market are tested. Through the experiment, we show that the international standard fails to provide a methodology which gives control limits for EMC when the electronic products in question are produced. We feel that an improved EMC control procedure presented here can better meet the needs of radiated emission control.

Much health related research depends heavily on the analysis of a rapidly expanding universe of observational data. A challenge in analysis of such data is the lack of sound statistical methods and tools that can address multiple facets of estimating treatment or exposure effects in observational studies with a large number of covariates. We sought to advance methods to improve analysis of large observational datasets with an end goal of understanding the effect of treatments or exposures on health. First we compared existing methods for propensity score (PS) adjustment, specifically Bayesian propensity scores. This concept had previously been introduced (McCandless et al., 2009) but no rigorous evaluation had been done to evaluate the impact of feedback when fitting the joint likelihood for both the PS and outcome models. We determined that unless specific steps were taken to mitigate the impact of feedback, it has the potential to distort estimates of the treatment effect. Next, we developed a method for accounting for uncertainty in confounding adjustment in the context of multiple exposures. Our method allows us to select confounders based on their association with the joint exposure and the outcome while also accounting for the uncertainty in the confounding adjustment. Finally, we developed two methods to combine het- erogenous sources of data for effect estimation, specifically information coming from a primary data source that provides information for treatments, outcomes, and a limited set of measured confounders on a large number of people and smaller supplementary data sources containing a much richer set of covariates. Our methods avoid the need to specify the full joint distribution of all covariates.

BACKGROUND: Morbidity, defined as disease history, is an important and well-known confounder in epidemiologic studies. Numerous methods have been developed over the last 30 years to measure morbidity via valid and reliable processes. OBJECTIVE: The goal of the current study was to evaluate, via comparative predictive validity assessment, the Centers for Medicaid and Medicare Studies Hierarchical Condition Category (CMS-HCC) comorbidity model for its ability to improve the prediction of 12-month all-cause mortality among a Medicare population compared to previously published comorbidity index models. There were three specific aims: (1) challenge the current state of risk adjustment among aged populations via an evaluation of the comparative predictive validity of one novel and four existing models to predict all-cause mortality within 12 months among a heterogeneous population of Medicare beneficiaries; (2) Investigate the comparative predictive validity of the five models to predict all-cause mortality within 12 months among two homogenous populations diagnosed with ischemic heart disease and selected cancers, including prostate cancer, lung cancer, colorectal cancer, breast cancer, pancreas cancer, and endometrial cancer; and (3) measure each comorbidity model's ability to control for a known example of confounding by indication. METHODS: A retrospective cohort design was used for all specific aims. Study 1 included 257,641 Medicare beneficiaries enrolled in three Medicare Advantage prescription drug health plans in Alabama, Florida, or Ohio in 2010 and 2011. Study 2 limited analysis to 14,260 and 66,440 beneficiaries with administrative evidence of selected cancers or ischemic heart disease in 2010, respectively. Study 3 limited analysis to the beneficiaries with ischemic heart disease. For each participant, comorbidity risk scores for the following five models were generated using administrative data from 2010: an age/sex model, the Romano adaption of the Charlson Comorbidity Index (CCI) model, the Putnam adaptation of the Chronic Disease Score Model (CDS), the CMS version of the Hierarchical Condition Category (CMS-HCC) model, and the Agency for Healthcare Research and Quality (AHRQ) adaptation of the Elixhauser model. The prospective predictive validity of the models to predict all-cause mortality during 2011 was compared via the c statistic test. Participants with ischemic heart disease were randomly allocated retrospectively to either 1) a group that had "received" a hypothetical "Drug A" in 2010 or 2) a group that had "received" a hypothetical "Drug B" in 2010. In order to evaluate the impact of confounding by indication, a weighting factor was applied to the randomization process in order to force the 33,220 participants randomized to "Drug A" to have a 2.736 times higher likelihood of having at least one acute inpatient hospitalization in 2010. Each comorbidity model's ability to control for the contrived confounding by indication was evaluated via relative risk of death. RESULTS: The CMS-HCC model had statistically significant higher c-statistic values than all four existing comorbidity indices among the heterogeneous Medicare Advantage population (N=257,641) and the homogeneous populations with breast cancer (N=4,160) and prostate cancer (N=6,594). The CMS-HCC model displayed similar performance for lung cancer (N=1,384), colorectal cancer (N=1,738), endometrial cancer (N=232), and ischemic heart disease (N=66,640) and statistically significant lower performance for pancreas cancer (N=152). The log-transformed CMS-HCC model was the only model to generate a non-significant association between exposure to "Drug A" and subsequent mortality. CONCLUSION: In general, the CMS-HCC model is the preferred comorbidity measure due to its predictive performance. However, other comorbidity models may be optimal for diseases with low prevalence and/or high mortality. Researchers should carefully and thoughtfully select a comorbidity model to assess the existence and direction of confounding. The CMS-HCC model should be log-transformed when used as a dependent variable since the score is a ratio level measurement that displays a normal distribution when log transformed. The resulting score is less likely to violate the assumptions (i.e. violations of normality) of common statistical models due to extreme values. The national availability of CMS-HCC scores for all Medicare beneficiaries provides researchers with access to a new tool to measure co-morbidity among older Americans using an empirically weighted, single score. In terms of policy, it is recommended that CMS produce CMS-HCC scores for all Medicare beneficiaries on a rolling 12 month basis for each month during the year. The availability of monthly scores would increase the ease of use of the score, as well as help facilitate more rapid adoption of the tool.

This thesis develops techniques for adjusting for selection bias using Gaussian process models. Selection bias is a key issue both in sample surveys and in observational studies for causal inference. Despite recently emerged techniques for dealing with selection bias in high-dimensional or complex situations, use of Gaussian process models and Bayesian hierarchical models in general has not been explored. Three approaches are developed for using Gaussian process models to estimate the population mean of a response variable with binary selection mechanism. The first approach models only the response with the selection probability being ignored. The second approach incorporates the selection probability when modeling the response using dependent Gaussian process priors. The third approach uses the selection probability as an additional covariate when modeling the response. The third approach requires knowledge of the selection probability, while the second approach can be used even when the selection probability is not available. In addition to these Gaussian process approaches, a new version of the Horvitz-Thompson estimator is also developed, which follows the conditionality principle and relates to importance sampling for Monte Carlo simulations. Simulation studies and the analysis of an example due to Kang and Schafer show that the Gaussian process approaches that consider the selection probability are able to not only correct selection bias effectively, but also control the sampling errors well, and therefore can often provide more efficient estimates than the methods tested that are not based on Gaussian process models, in both simple and complex situations. Even the Gaussian process approach that ignores the selection probability often, though not always, performs well when some selection bias is present. These results demonstrate the strength of Gaussian process models in dealing with selection bias, especially in high-dimensional or complex situations. These results also demonstrate that Gaussian process models can be implemented rather effectively so that the benefits of using Gaussian process models can be realized in practice, contrary to the common belief that highly flexible models are too complex to use practically for dealing with selection bias.

Conselho Nacional de Desenvolvimento Científico e Tecnológico / Apesar do crescente reconhecimento do potencial dos diagramas causais por epidemiologistas, essa técnica ainda é pouco utilizada na investigação epidemiológica. Uma das possíveis razões é que muitos temas de investigação exigem modelos causais complexos. Neste trabalho, a relação entre estresse ocupacional e obesidade é utilizada como um exemplo de aplicação de diagramas causais em questões relacionadas a confundimento. São apresentadas etapas da utilização dos diagramas causais, incluindo a construção do gráfico acíclico direcionado, seleção de variáveis para ajuste estatístico e a derivação das implicações estatísticas de um diagrama causal. A principal vantagem dos diagramas causais é tornar explícitas as hipóteses adjacentes ao modelo considerado, permitindo que suas implicações possam ser analisadas criticamente, facilitando, desta forma, a identificação de possíveis fontes de viés e incerteza nos resultados de um estudo epidemiológico. / Despite the increasing recognition of the potential of causal diagrams by epidemiologists, this technique has not been widely used in epidemiological research. One possible reason is that many research topics require complex causal models. In this article, the relationship between occupational stress and obesity is used as an example of application of causal diagrams on confounding. Some steps are presented, including the construction of the directed acyclic graph, the selection of variables for statistical control and the derivation of the statistical implications of a causal diagram. The main advantage of causal diagrams is to make the assumptions explicit, thus facilitating critical evaluations and the identification of possible sources of bias and uncertainty in the results of an epidemiological study.

abstract: No-confounding designs (NC) in 16 runs for 6, 7, and 8 factors are non-regular fractional factorial designs that have been suggested as attractive alternatives to the regular minimum aberration resolution IV designs because they do not completely confound any two-factor interactions with each other. These designs allow for potential estimation of main effects and a few two-factor interactions without the need for follow-up experimentation. Analysis methods for non-regular designs is an area of ongoing research, because standard variable selection techniques such as stepwise regression may not always be the best approach. The current work investigates the use of the Dantzig selector for analyzing no-confounding designs. Through a series of examples it shows that this technique is very effective for identifying the set of active factors in no-confounding designs when there are three of four active main effects and up to two active two-factor interactions. To evaluate the performance of Dantzig selector, a simulation study was conducted and the results based on the percentage of type II errors are analyzed. Also, another alternative for 6 factor NC design, called the Alternate No-confounding design in six factors is introduced in this study. The performance of this Alternate NC design in 6 factors is then evaluated by using Dantzig selector as an analysis method. Lastly, a section is dedicated to comparing the performance of NC-6 and Alternate NC-6 designs. / Dissertation/Thesis / Masters Thesis Industrial Engineering 2014

Conselho Nacional de Desenvolvimento Científico e Tecnológico / Apesar do crescente reconhecimento do potencial dos diagramas causais por epidemiologistas, essa técnica ainda é pouco utilizada na investigação epidemiológica. Uma das possíveis razões é que muitos temas de investigação exigem modelos causais complexos. Neste trabalho, a relação entre estresse ocupacional e obesidade é utilizada como um exemplo de aplicação de diagramas causais em questões relacionadas a confundimento. São apresentadas etapas da utilização dos diagramas causais, incluindo a construção do gráfico acíclico direcionado, seleção de variáveis para ajuste estatístico e a derivação das implicações estatísticas de um diagrama causal. A principal vantagem dos diagramas causais é tornar explícitas as hipóteses adjacentes ao modelo considerado, permitindo que suas implicações possam ser analisadas criticamente, facilitando, desta forma, a identificação de possíveis fontes de viés e incerteza nos resultados de um estudo epidemiológico. / Despite the increasing recognition of the potential of causal diagrams by epidemiologists, this technique has not been widely used in epidemiological research. One possible reason is that many research topics require complex causal models. In this article, the relationship between occupational stress and obesity is used as an example of application of causal diagrams on confounding. Some steps are presented, including the construction of the directed acyclic graph, the selection of variables for statistical control and the derivation of the statistical implications of a causal diagram. The main advantage of causal diagrams is to make the assumptions explicit, thus facilitating critical evaluations and the identification of possible sources of bias and uncertainty in the results of an epidemiological study.