Дисертації з теми "Confounding Factor"

Musculoskeletal ultrasound (US) is an imaging modality with the potential to revolutionise the way we practise rheumatology. Ultrasound has been shown to be more sensitive and valid when compared to conventional imaging methods and clinical examination in the assessment of inflammatory arthropathies (lA). However, major limitations to its widespread use have been its perceived poor reliability and unknown diagnostic discriminative ability. Image acquisition is a key point for improving reliability. A systematic literature review was performed and a survey of experts was obtained in order to understand the important confounding factors in image acquisition. The effect of two of these factors on synovitis detection was then examined. Non-steroidal anti-inflammatory drugs (NSAIDs) usage was found to significantly mask both grey scale (GS) and power Doppler (PD) signal and resultedinlower assessment of disease activity. Change in the joint position during scanning was found to significantly alter US findings in patients with lA. Scanning the hands in a flat position and the knees in a 30° position provided the highest GS and PD scores suggesting that they may be the optimal scanning positions. The ability of US to discriminate rheumatoid arthritis (RA) from other diseases by detecting bone erosions was explored by examining the specificity of US detected bone erosion. The overall presence of US detected erosions was not a specific fmding for RA. However, erosions in certain anatomical sites were specific for RA. In conclusion, this thesis has demonstrated that certain factors. including concurrent NSAID intake and joint position can affect US image acquisition. Ultrasound does / IV have the ability to discriminate RA from other diseases by specifically detecting bone erosions in target joints. Standardisation of the US confounding factors and better understanding of the specificity of common US findings may enhance the role of US in the assessment of IA.

Abstract To date, most epidemiological studies have shown a weak or moderate association between dental diseases such as periodontal infections, dental caries and tooth loss, and atherosclerotic vascular diseases. However, the nature of this association is not known; it may be due to the biological effect of oral infections on initiation or progress of atherosclerosis or it may be non-causal due to determinants in common, either biological or behavioural. Methodological shortcomings, inconsistent results and a lack of definite proof from intervention studies have led to the conclusion that causality between dental diseases and atherosclerotic vascular diseases has not been established. The aim of this study was to produce evidence on the nature of the association between dental diseases and atherosclerotic vascular diseases. The study uses data from the 1966 Birth Cohort of Northern Finland (N = 11,637). The data were collected in 1997–1998, when the cohort members had reached 31 years of age. The respondents were asked through a postal questionnaire about their oral health. In addition, respondents were asked about their general health and oral and general health habits. The response rate was 75.3%. Those who lived in Northern Finland or the capital city region were invited to clinical health examination (N = 8,463). Altogether 5,696 subjects supplied the data, representing 67.3% of those who were invited to the clinical examination. While the study showed an association of self-reported gingivitis, dental caries and tooth loss with the prevalent angina pectoris, it also showed that these self-reported dental diseases were not important determinants for elevated C-reactive protein levels. This suggests that the associations that were found between dental conditions and prevalent angina pectoris are mainly caused by factors other than biological mechanisms related to infection or inflammation. The lack of a biological explanation related to infections or inflammatory processes suggests that other biological mechanisms or biases, including confounding, should be considered as an alternative explanation. However, it must be noted that the possibility that oral infections also contribute to the development of atherosclerosis should not be rejected either.

Répondre à une question de recherche causale dans un contexte d’étude observationnelle nécessite desélectionner des variables de confusion. Leur intégration dans un modèle multivarié en tant que co-variablespermet de diminuer le biais dans l’estimation de l'effet causal de l'exposition sur le critère de jugement. Leuridentification est réalisée grâce à des diagrammes causaux (DCs) ou des graphes orientés acycliques. Cesreprésentations, composées de noeuds et d'arcs orientés, permettent d’éviter la sélection de variables quiaugmenteraient le biais, comme les variables de médiation et de collision. Les méthodes existantes deconstruction de DCs manquent cependant de systématisme et leur représentation de formalisme, d’expressivité etde complétude. Afin de proposer un cadre de construction formel et complet capable de représenter toutes lesinformations nécessaires à la sélection des variables sur un DC enrichi, d’analyser ce DC et surtout d’expliquerles résultats de cette analyse, nous avons proposé d'utiliser un modèle ontologique enrichi de règles d'inférences.Un modèle ontologique permet notamment de représenter les connaissances sous la forme de graphe expressif etformel composé de classes et de relations similaires aux noeuds et arcs des DCs. Nous avons développél’ontologie OntoBioStat (OBS) à partir d’une liste de questions de compétence liée à la sélection des variables etde l'analyse de la littérature scientifique relative aux DCs et aux ontologies. Le cadre de construction d’OBS estplus riche que celui d’un DC, intégrant des éléments implicites tels que les causes nécessaires, contextuels d’uneétude, sur l’incertitude de la connaissance et sur la qualité du jeu de données correspondant. Afin d’évaluerl’apport d’OBS, nous l’avons utilisée pour représenter les variables d’une étude observationnelle publiée etavons confronté ses conclusions à celle d’un DC. OBS a permis d'identifier de nouvelles variables de confusiongrâce au cadre de construction différent des DCs et aux axiomes et règles d'inférence. OBS a également étéutilisée pour représenter une étude rétrospective en cours d’analyse : le modèle a permis d’expliquer dans unpremier temps les corrélations statistiques retrouvées entre les variables de l’étude puis de mettre en évidence lespotentielles variables de confusion et leurs éventuels substituts ("proxys"). Les informations sur la qualité desdonnées et l’incertitude des relations causales ont quant à elles facilité la proposition des analyses de sensibilité,augmentant la robustesse de la conclusion de l’étude. Enfin, les inférences ont été expliquées grâce aux capacitésde raisonnement offertes par le formalisme de représentation d'OBS. À terme OBS sera intégrée dans des outilsd’analyse statistique afin de bénéficier des bibliothèques existantes pour la sélection des variables et de permettreson utilisation par les épidémiologistes et les biostatisticiens
Responding to a causal research question in the context of observational studies requires the selection ofconfounding variables. Integrating them into a multivariate model as co-variables helps reduce bias in estimatingthe true causal effect of exposure on the outcome. Identification is achieved through causal diagrams (CDs) ordirected acyclic graphs (DAGs). These representations, composed of nodes and directed arcs, prevent theselection of variables that would introduce bias, such as mediating and colliding variables. However, existingmethods for constructing CDs lack systematic approaches and exhibit limitations in terms of formalism,expressiveness, and completeness. To offer a formal and comprehensive framework capable of representing allnecessary information for variable selection on an enriched CD, analyzing this CD, and, most importantly,explaining the analysis results, we propose utilizing an ontological model enriched with inference rules. Anontological model allows for representing knowledge in the form of an expressive and formal graph consisting ofclasses and relations similar to the nodes and arcs of Cds. We developed the OntoBioStat (OBS) ontology basedon a list of competency questions about variable selection and an analysis of scientific literature on CDs andontologies. The construction framework of OBS is richer than that of a CD, incorporating implicit elements likenecessary causes, study context, uncertainty in knowledge, and data quality. To evaluate the contribution of OBS,we used it to represent variables from a published observational study and compared its conclusions with thoseof a CD. OBS identified new confounding variables due to its different construction framework and the axiomsand inference rules. OBS was also used to represent an ongoing retrospective study analysis. The modelexplained statistical correlations found between study variables and highlighted potential confounding variablesand their possible substitutes (proxies). Information on data quality and causal relation uncertainty facilitatedproposing sensitivity analyses, enhancing the study's conclusion robustness. Finally, inferences were explainedthrough the reasoning capabilities provided by OBS's formal representation. Ultimately, OBS will be integratedinto statistical analysis tools to leverage existing libraries for variable selection, making it accessible toepidemiologists and biostatisticians

Record numbers of new diagnoses of HIV infection have been recorded in the UK in recent years. However, whether these are historic infections now being diagnosed or evidence of ongoing HIV transmission is unclear from these data. The Serological Testing Algorithm for Recent HIV Seroconversion (STARHS) is a generic term for .a number of different laboratory techniques that can be used to distinguish recent HIV infections occurring in the 4-6 months prior to sampling from long standing HIV infections. When these data are analysed with appropriate demographic data it is possible to estimate the rate of acquisition ofHIV infection or incidence . . However the method has many .limitations. This thesis identifies and examines the confounding factors that limit the applications of the STARHS technologies or alter its accuracy. It quantifies the degree of misclassification of specimens as recent HIV infections due to use of effective anti-retroviral therapy in patients and proposes uses of ST ARHS data in populations where multiple HIV subtypes circulate. and incidence estimates are difficult to determine. Having excluded confounding factors, the STARHS 'detuned' appr~ach is useq to determine HIV Incidence in men who·have sex with men attending Sexually Transmitted Infection (STI) clinics in the UK as part of an unlinked anonymous HIV prevalence monitoring programme. This demonstrated that ongoing transmission of HIV is occurring in this population and that despite widespread use of anti-retroviral therapies the rate of HIV transmission has shown no decline. This thesis adds new insight to and understanding of the complex mechanisms that limit the application of laboratory techniques for identifying recent HIV infection. Furthermore, it demonstrates that the STARHS teclmique, when used appropriately, is able to provide reliable and sensitive estimates of HIV incidence, thus improving understanding of recent trends in the HIV epidemic.

With an increase in life expectancy, the number of older people affected by cognitive decline and dementia is rising, causing major, global public health concerns. However, there is substantial variation in the rate and magnitude of cognitive decline experienced among ageing individuals. Evidence suggests that many age-associated changes in cognitive functioning can be explained by modifiable lifestyle factors such as smoking, physical activity and diet choices. The weight of the evidence supports the promotion of a healthy lifestyle as an effective strategy for healthy cognitive ageing. Many epidemiological studies have drawn causal conclusions with regard to the positive and direct benefits of lifestyle, yet few have considered the possible confounding role of prior cognitive ability in explaining the lifestyle and cognition relationship in older age. Given the potential for reverse causation, whereby better prior cognitive functioning leads to a greater uptake of healthy behaviours rather than vice versa, it is a mechanism which should be studied, but rarely is. The present thesis focuses on the possible confounding effect of prior cognitive ability on the cross-sectional relationships between lifestyle factors and cognitive ability domains in later-life. The core of the thesis is a series of independent, peer-reviewed (six first-author and one co-author) journal articles in the public domain. Data were derived from the Lothian Birth Cohort 1936 study (n = 1091), a sample of relatively healthy, community-dwelling men and women aged 70 years from Edinburgh, Scotland, for whom childhood (age 11) mental test scores are available. The lifestyle factors investigated were caffeine consumption, alcohol consumption, dietary patterns, body mass index, smoking, serum cholesterol, and physical activity. Cognitive function was assessed across five major ageing-related domains: age 70 IQ (based on the same test that was taken in childhood), general cognitive ability (g), processing speed, memory, and verbal ability. General linear models (ANCOVA) were adjusted for the following covariates: age; sex; childhood cognitive ability; and socioeconomic status (SES). Other potential covariates were additionally adjusted for as necessary. Overall, the positive and significant associations observed between ‘healthy’ lifestyle factors and better cognitive functions at age 70 were consistent with previous research; their effect size was around 1% of the variance in cognitive tests scores. However, these relationships were markedly attenuated (by on average 80%) by a higher childhood cognitive ability and adult SES; for the most part, associations were reduced to non-significance. None of the lifestyle factors were consistent predictors of performance across cognitive domains, though smoking avoidance, a physically active lifestyle, and moderate intake of alcohol, appeared to have the most potential. The key novel finding of this thesis is that, in addition to having predictive value for lifestyle choices over 60 years later, cognitive ability at age 11 accounted for the majority of the cross-sectional associations between lifestyle factors and cognitive abilities in later-life. This finding is consistent with the theory of confounding or even reverse causation. That is, individuals with higher lifetime ‘trait’ cognitive ability may be more likely to adopt a lifestyle which protects against cognitive decline. Rather than a unidirectional or indirect effect of health behaviours on cognitive function, the present findings suggest there may be a dynamic cycle involving cognition, self-management of health and ultimate cognitive outcomes.

BACKGROUND: Fish consumption and intake of omega-3 fatty acids from fish are associated with a lower risk of cardiovascular disease. However, a prospective study from northern Sweden showed that high consumption of fish is associated with an increased risk of stroke in men, but not in women. The current study aimed to determine if fish consumption is differently related to lifestyle in men compared with women in northern Sweden. METHODS: Lifestyle information on 32,782 men and 34,866 women (aged 30--60 years) was collected between 1992 and 2006 within the Vasterbotten Intervention Programme (a health intervention in northern Sweden). Spearman correlation coefficients (Rs) were calculated for associations between self-reported consumption of fish and other food items or lifestyle variables. RESULTS: Fish consumption was positively associated with other foods considered healthy (e.g., root vegetables, lettuce/cabbage/spinach/broccoli, chicken, and berries; Rs = 0.21-0.30), as well as with other healthy lifestyle factors (e.g., exercise and not smoking) and a higher educational level, in both men and women. The only gender difference found, concerned the association between fish consumption and alcohol consumption. Men who were high consumers of fish had a higher intake of all types of alcohol compared with low to moderate fish consumers. For women, this was true only for wine. CONCLUSIONS: Except for alcohol, the association between fish consumption and healthy lifestyle did not differ between men and women in northern Sweden. It is important to adjust for other lifestyle variables and socioeconomic variables in studies concerning the effect of fish consumption on disease outcome.

Includes abstract.
Includes bibliographical references.
The general aim of this thesis is therefore to analyse alcohol consumption variables and possible confounding risk factors associated with risky sexual behaviour amongst women in the urban city of Tshwane in Gauteng and the rural Western Cape sites, respectively.

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

Les études observationnelles en pharmacoépidémiologie sont souvent mises en place pour évaluer un médicament mis sur le marché récemment ou concurrencé par de nombreuses alternatives thérapeutiques. Cette situation conduit à devoir évaluer l'effet d'un médicament dans une cohorte comprenant peu de sujets traités, c'est à dire une population où l'exposition d'intérêt est rare. Afin de prendre en compte les facteurs de confusion dans cette situation, certains auteurs déconseillent l'utilisation du score de propension au profit du score pronostique, mais cette recommandation ne s'appuie sur aucune étude évaluant spécifiquement les faibles prévalences de l'exposition, et ignore le type d'estimation, conditionnelle ou marginale, fournie par chaque méthode d'utilisation du score pronostique.La première partie de ce travail évalue les méthodes basées sur le score de propension pour l'estimation d'un effet marginal en situation d'exposition rare. La deuxième partie évalue les performances des méthodes basées sur le score pronostique rapportées dans la littérature, introduit de nouvelles méthodes basées sur le score pronostique adaptées à l'estimation d'effets conditionnels ou marginaux, et les compare aux performances des méthodes basées sur le score de propension. La dernière partie traite des estimateurs de la variance des effets du traitement. Nous présentons les conséquences liées à la non prise en compte de l'étape d'estimation du score de propension et du score pronostique dans le calcul de la variance. Nous proposons et évaluons de nouveaux estimateurs tenant compte de cette étape
Pharmacoepidemiologic observational studies are often conducted to evaluate newly marketed drugs or drugs in competition with many alternatives. In such cohort studies, the exposure of interest is rare. To take into account confounding factors in such settings, some authors advise against the use of the propensity score in favor of the prognostic score, but this recommendation is not supported by any study especially focused on infrequent exposures and ignores the type of estimation provided by each prognostic score-based method.The first part of this work evaluates the use of propensity score-based methods to estimate the marginal effect of a rare exposure. The second part evaluates the performance of the prognostic score based methods already reported in the literature, compares them with the propensity score based methods, and introduces some new prognostic score-based methods intended to estimate conditional or marginal effects. The last part deals with variance estimators of the treatment effect. We present the opposite consequences of ignoring the estimation step of the propensity score and the prognostic score. We show some new variance estimators accounting for this step

Pour faire face à des perturbations environnementales, les organismes ont développé des défenses antitoxiques couramment utilisés comme biomarqueurs dans l'évaluation de la qualité des milieux. Cependant, de nombreux facteurs confondants comme la température ou le genre, influencent la réponse de ces biomarqueurs. Il est ainsi nécessaire de connaitre les effets de ces facteurs afin d'imputer correctement la réponse biologique mesurée à la présence de polluants. Dans ce contexte, nous avons choisi d'étudier l'influence d'un parasite acanthocéphale, Polymorphus minutus, transmis horizontalement et de microsporidies à transmission verticale, sur les réponses physiologiques de leur hôte, Gammarus roeseli, un crustacé amphipode d'eau douce d'intérêt écotoxicologique. Pour cela, nous avons choisi d'étudier le glutathion, tripeptide jouant un rôle central dans les systèmes antitoxiques, sa synthèse (i.e. activité gamma-glutamylcystéine ligase), les réserves énergétiques (i.e. lipides, glycogène, protéines) et un biomarqueur d'effets toxiques, le malondialdéhyde. L'influence du parasitisme a été appréhendé dans différents cas d'études: (i) chez G. roeseli infecté par P. minutus, (ii) chez G. roeseli infecté par des microsporidies (Dictyocoela roeselum essentiellement) et (iii) chez G. roeseli co-infecté par ces deux types de parasite. Chaque cas d'étude a été réalisé en absence de stress et lors d'une exposition au cadmium. Nous avons mis en évidence qu'en absence de contamination, la présence de P. minutus et une co-infection par P. minutus et D. roeselum affectent les biomarqueurs de G. roeseli. Après exposition au cadmium, la présence de parasites (i.e. infection simple ou co-infection) modifie la mobilisation des défenses antitoxiques et accentue les effets toxiques subits par l'hôte. Les résultats obtenus au cours de ce travail mettent en avant le caractère confondant du parasitisme en écotoxicologie et souligne l'importance de prendre en compte ce paramètre lors de l'évaluation de la qualité des milieux
To cope with environmental disturbances, organisms have developed antitoxic defenses commonly used as biomarkers in environmental risk assessment. However, many confounding factors such as temperature and gender could influence biomarker responses. It seems hence necessary to investigate their effects, in order to attribute biological responses only to pollutants. In this context, we investigated the influence of parasitism by studying the acanthocephalan parasite Polymorphus minutus, horizontally transmitted, and microsporidia parasites, vertically transmitted, on the physiological responses of their common host, the freshwater amphipod Gammarus roeseli, a classical model used in ecotoxicology. We investigated the glutathione, a tripeptide having a key role in antitoxic systems, its synthesis (i.e. gamma-glutamylcysteine ligase activity), energy reserves (i.e. lipids, glycogen, proteins) and a toxicity biomarker, the malondialdehyde. The influence of parasitism was considered in different studies: (i) in G. roeseli infected by P. minutus only, (ii) in G. roeseli infected by microsporidia (mainly Dictyocoela roeselum) and (iii) in G. roeseli coinfected by both parasites. Each study was carried out in absence of pollutants and under cadmium stress. We highlighted that, in the absence of contamination, only P. minutus and the co-infection affect the G. roeseli biomarker assessments. After cadmium exposure, the presence of parasites (i.e. single infection or co-infection) influences the mobilization of antitoxic defences, and accentuates toxic effects in their hosts. Our results underline the confounding nature of parasitism in ecotoxicology and thus, highlight the importance to take into account this parameter in the environmental risk assessment

Nous présentons des méthodes statistiques reposant sur des problèmes de factorisation matricielle. Une première méthode permet l'inférence rapide de la structure de populations à partir de données génétiques en incluant l'information de proximité géographique. Une deuxième méthode permet de corriger les études d'association pour les facteurs de confusion. Nous présentons dans ce manuscrit les modèles, ainsi que les aspects théoriques des algorithmes d'inférence. De plus, à l'aide de simulations numériques, nous comparons les performances de nos méthodes à celles des méthodes existantes. Enfin, nous utilisons nos méthodes sur des données biologiques réelles. Nos méthodes ont été implémentées et distribuées sous la forme de packages R : tess3r et lfmm
We present statistical methods based on matrix factorization problems. A first method allows efficient inference of population structure from genetic data and including geographic proximity information. A second method corrects the association studies for confounding factors. We present in this manuscript the models, as well as the theoretical aspects of the inference algorithms. Moreover, using numerical simulations, we compare the performance of our methods with those of existing methods. Finally, we use our methods on real biological data. Our methods have been implemented and distributed as R packages: tess3r and lfmm

Les réglementations conduisent à multiplier les évaluations du risque écologique lié au rejet des substances chimiques dans les milieux aquatiques avec pour objectif de protéger les populations naturelles. Cependant, ce niveau de protection visé ne peut être utilisé facilement pour établir un lien direct entre une contamination et ses effets. Pour palier cette difficulté, les approches multi-échelles basées sur l'étude des effets des contaminations sur des marqueurs individuels puis sur l'extrapolation de ces effets au niveau des poipulations à partir de modèles de dynamique de population représentent des démarches prometteuses et commencent à être bien acceptées dans les démarches prédictives.Cependant, leur utilisation pour le diagnostic de la qualité des milieux aquatiques reste pour le moment très rare, d'une part en raison de la forte variabilité des réponses des marqueurs individuels in situ liée à l'influence de divers facteurs environnementaux confondants pour l'évaluation de la toxicité, et d'autre part à cause du manque de pertinence environnementale des modèles actuellement proposés. Centré sur l'utilisation de deux espèces couramment observées dans les cours d'eau européens et présentant des caractéristiques écologiques et phylogéniques contrastées : le crustacé Gammarus fossarum et le mollusque Potamopyrgus antipodarum, ce travail doctoral a permis : 1- de proposer une méthodologie pour prendre en compte l'influence des facteurs de confusion dans le but d'améliorer la lecture des bioessais in situ basés sur la mesure des marqueurs individuels sur des organismes encagés et 2- de développer des modèles de dynamique de population écologiquemnt pertinents
The regulatory framework lead to increase the assessments of the ecological risk linked to the dischaarge of chemical substances in aquatic environment with the aim to protect natural populations. However, this target level of protection cannot be used so easily to etablish a direcet link between a contamination and its effects. In order to overcome this difficulty, the multi-scale approaches based on the study of the effects of the contaminations on individual markers and then on the extrapolation of these effects at the population level with population dynamic models reprensent promising tools and start to bewell accepted in predictive processes. Yet, their use for the diagnosis of water quality remainsrare for the moment, on the one hand, because of the important varaibility of answers of individual makers in situ linked to the influence of diverse confounding environmental factors for the assessments of toxicity and, on the other hand, because pf the lack of environmental relevance of models currently proposed. Focused on the use of 2 species widely observed in European rivers and presenting contrastingecological and phylogenetic characteristics : the Grammar fossarum crutacean and the Potamopyrgus antipodarum mollusc, this doctoral degree first propose a methodology to consider the influenceof confounding factors in order to improve the reading of biological in situ tests based on the measurment of individual markers on caged organisms and second, to develop ecologically relevant population models. Thus, this work allowed to underline the importance of the consideration of confounding factors (i.e. temperature), in different in situ tests based on the measurment of individul markers on caged organisms. Moreover, population models for both species have been defined in order to test the influence of life histories and seasonal variations on demographic sensitivity of populations

Osteoporosis is a significant public health problem in the U.S. It not only affects the physical well-being of the older women but also creates a substantial financial burden for the health care system. The mainstay of osteoporosis medications is bisphosphonate treatment of which alendronate and risedronate are the most commonly prescribed in clinical practice. However, there have been no head-to-head randomized controlled trials (RCTs) evaluating the effects of these two bisphosphonates on fracture outcomes. In the absence of RCTs, observational studies are necessary to provide alternative evidence on the comparative effectiveness between alendronate and risedronate on fracture outcomes. However, existing observational studies have provided inconclusive results partially due to residual confounding from unobserved variables such as patients’ health status or behavior. IV analysis may be one method to address unmeasured confounding bias in observational studies. While it has not been applied in bisphosphonate research, it has been used in research on a variety of other prescription medications. In this dissertation, we applied the IV approach with an IV, date of generic alendronate availability, to evaluate the comparative effectiveness between alendronate and risedronate using observational data. This dissertation improved current research in several ways. First, we extended the IV approach to research on bisphosphonates. Second, compared with the current observational studies on bisphosphonates, this dissertation may more accurately estimate the relative effects between alendronate and risedronate because IV analysis is not subject to unmeasured confounding bias. Third, the study results extended the current evidence of the comparative effectiveness between the two most commonly prescribed bisphosphonates. Finally, we proposed and provided empirical evidence of a new IV that might be used for future prescription drug research. The finding of this dissertation can be summarized from three aspects. First, we found that the evidence supported the validity of the date of generic availability as an IV in the study of bisphosphonates. Second, applying IV approach to study the comparative effectiveness of alendronate and risedronate, we found that alendronate and risedronate were comparable to reduce the risk of 12-month non-vertebral fractures in older women. Since generic alendronate is availability on the market while generic risedronate is not, promoting the use of alendronate may help reduce the healthcare cost and not sacrifice the clinical effectiveness. Finally, by comparing the proposed IV with a popular IV-physician preference, we found that both the calendar time IV based on the date of generic availability and the physician preference appeared to be valid. It might be practically easier to use the calendar time IV than the physician preference IV.

Les effets du changement climatique sur la biodiversité à l'échelle du globe sont maintenant sans équivoque. Parmi les écosystèmes affectés, les cours d'eau sont particulièrement vulnérables aux fluctuations du climat. Les modifications de structure et de composition des communautés aquatiques constituent alors un signal intégrateur des réponses écologiques à ces changements climatiques. Dans certaines circonstances, ce signal peut également être exacerbé, modéré ou potentiellement masqué par d'autres variations abiotiques.L'objectif de ce travail de thèse était donc d'évaluer l'effet relatif de différents facteurs de forçage sur les tendances d'évolution à long-terme des communautés de macroinvertébrés benthiques de grands cours d'eau, dans un contexte de changement climatique global. Nos investigations ont ciblé plus spécifiquement la Loire moyenne car, en tant que grande rivière de plaine, elle est particulièrement concernée par les problèmes liés aux stress multiples. En ce qui concerne le choix du compartiment biologique, la faune benthique a été sélectionnée pour son intérêt reconnu en bioévaluation des cours d'eau, notamment grâce à une grande diversité de réponses potentielles aux perturbations environnementales et à un certain nombre d'avantages méthodologiques.Dans un premier temps, nous nous sommes appliqués à caractériser les modifications temporelles de dix paramètres abiotiques majeurs sur trois décennies (1977-2008). Nous avons notamment mis en évidence le réchauffement graduel de la Loire (c. +1,2°C en moyenne annuelle sur les trois décades), couplé à une réduction significative du débit moyen (c. -25%), les deux tendances étant exacerbées en période chaude (mai-août). Dans le même temps, alors que ces modifications étaient susceptibles de provoquer une augmentation du niveau trophique du fleuve, nous avons souligné l'existence d'un effet confondant de l'amélioration des traitements d'épuration (i.e. réduction des apports en phosphore), se traduisant principalement par la baisse des concentrations en phosphates et la limitation des organismes phytoplanctoniques.Dans un second temps, nous nous sommes intéressés aux réponses structurelles et fonctionnelles long-terme (i.e. 30 ans) des invertébrés de la Loire à ces deux évolutions concomitantes majeures. En premier lieu, nous avons constaté que le réchauffement et, dans une moindre mesure, la réduction du débit sont impliqués dans la disparition ou le déclin progressif de taxons rhéophiles et psychrophiles (e.g. Chloroperlidae). En parallèle, ces modifications hydroclimatiques expliquent une majeure partie de l'apparition et de la colonisation de taxons limnophiles et thermophiles, comprenant certaines espèces invasives (e.g. Corbicula sp.). En termes de réponses fonctionnelles, ces changements s'illustrent par une évolution adaptative des profils de traits biologiques cohérente avec les prédictions a priori qu'il est possible de faire dans un cadre théorique. Toutefois, cette dérive progressive vers un assemblage généraliste et polluotolérant est partiellement confondue par l'amélioration de la qualité d'eau, qui explique l'apparition de nouveaux taxons polluo-sensibles lors des dernières années (e.g. Philopotamidae). Bien qu'aucune résilience taxonomique ou fonctionnelle ne soit constatée, cette tendance se manifeste, via les traits biologiques, par un transfert probable des ressources trophiques depuis les organismes phytoplanctoniques vers des biofilms épibenthiques et des macrophytes.Pour finir, nous avons exploré la variabilité spatiale multi-sites de ces tendances biotiques et abiotiques et montré que l'ensemble de ces résultats, qui corrobore déjà un certain nombre d'observations en Europe, constitue également un patron homogène le long de la Loire moyenne (sur près de 300 km) et entre différents hydrosystèmes (e.g. Loire, Seine, Meuse).

abstract: During the initial stages of experimentation, there are usually a large number of factors to be investigated. Fractional factorial (2^(k-p)) designs are particularly useful during this initial phase of experimental work. These experiments often referred to as screening experiments help reduce the large number of factors to a smaller set. The 16 run regular fractional factorial designs for six, seven and eight factors are in common usage. These designs allow clear estimation of all main effects when the three-factor and higher order interactions are negligible, but all two-factor interactions are aliased with each other making estimation of these effects problematic without additional runs. Alternatively, certain nonregular designs called no-confounding (NC) designs by Jones and Montgomery (Jones & Montgomery, Alternatives to resolution IV screening designs in 16 runs, 2010) partially confound the main effects with the two-factor interactions but do not completely confound any two-factor interactions with each other. The NC designs are useful for independently estimating main effects and two-factor interactions without additional runs. While several methods have been suggested for the analysis of data from nonregular designs, stepwise regression is familiar to practitioners, available in commercial software, and is widely used in practice. Given that an NC design has been run, the performance of stepwise regression for model selection is unknown. In this dissertation I present a comprehensive simulation study evaluating stepwise regression for analyzing both regular fractional factorial and NC designs. Next, the projection properties of the six, seven and eight factor NC designs are studied. Studying the projection properties of these designs allows the development of analysis methods to analyze these designs. Lastly the designs and projection properties of 9 to 14 factor NC designs onto three and four factors are presented. Certain recommendations are made on analysis methods for these designs as well.
Dissertation/Thesis
Ph.D. Industrial Engineering 2012

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.

碩士
高雄醫學大學
藥學研究所
98
Background The glycopeptide antibiotic vancomycin has been widely used in the treatment of gram-positive infectious diseases; especially those caused by methicillin-resistant Staphylococcus aureus (MRSA). In vancomycin therapy, plasma levels of vancomycin must be monitored to avoid side effects and obtain effective clinical responses in each individual. Several methods have been proposed for setting the initial dosage regimen of vancomycin therapy, including Moellering nomogram, Matzke nomogram and Rodvold etc.. By using the drug monitoring (TDM) data from adult MICU inpatients, this study had evaluated the appropriateness of the Matzke nomogram and tried to find confounding factor that influencing the vancomycin serum concemtration. Methods This is a single-center and retrospective study. All adult patients used vancomycin therapy and with TDM data from July 2004 to June 2009. Data was included in this population model dataset had been included in a previous population modelling study. Demographic data included patient age, gender, total body weight, serum creatinine (Scr) and serum concentrations were collected from TDM files. Creatinine clearance (CLcr) was calculated using the Cockcroft–Gault equation. Vancomycin using less than 48 hours or not using intravenous injection, pregnant or breast-feeding women, burn patients (>10%BSA) and patients on dialysis were excluded from the study. General demographic data with serum concentration and confounding factors were analyzed by linear regression analysis. Results The pharmacokinetics of vancomycin were characterized in 430 patients with different degrees of renal function after an initial dose of 24.8±1.1mg/kg and maintain dose 19.3 ± 1.1 mg/kg. According to the level of CLcr, there were 242, 162 and 26 patients subjects in CLcr > 60 mL/min (group I), 30~60 mL/min (group II) and 10~30 mL/min (group III), respectively. The results showed that peak level 80.2% patients in groups I, 71.6% patients in groups II and 88.5% patients in groups IIII were 20~40 mg/L; the trough level 62.0% patients in group I, 60.5% patients in groups II and 69.2% patients in groups IIII were 5~15 mg/L. When tested on Matzke method had the lowest area under the ROC curve, AUC 0.501 (0.392-0.611 95% C.I.) for peak level and AUC 0.517 (0.406-0.627 95% C.I.) for trough level, respeatively. The linear regression analysis showed that the ageing was the confounding favor in predicting vancomycin serum concentration. Conclusions Matzke nomogram was shown to achieve target plasma levels (5~15 mg/L) of vancomycin at a higher rate, but lack of convincing evidence of a correlation between serum concentrations and therapeutic outcome. Secondly, little comparative information is currently available as to the dosing of vancomycin in post-adolescent patient populations. This simulation indicated that the Matzke nomogram dosing interval was the most successful method for initial dose selection, but early serum concentration monitoring and adjustment of initial empirical and nomogram-derived doses is necessary.

La résonance magnétique cardiovasculaire sensible à l'oxygénation (OS-CMR) est devenue une modalité d'imagerie diagnostique pour la surveillance de changements dans l'oxygénation du myocarde. Cette technique offre un grand potentiel en tant qu'outil diagnostic primaire pour les maladies cardiovasculaires, en particulier la détection non-invasive d'ischémie. Par contre, il existe plusieurs facteurs potentiellement confondants de cette technique, quelques-uns d'ordre méthodologique comme les paramètres de séquençage et d'autres de nature physiologiques qui sont peut compris. En raison des effets causés par le contenu tissulaire d'eau, l'état d'hydratation peut avoir un impact sur l'intensité du signal. Ceci est un des aspects physiologiques en particulier dont nous voulions quantifier l'effet confondant par la manipulation de l'état d'hydratation chez des humains et l'observation des changements de l'intensité du signal dans des images OS-CMR. Méthodes: In vitro: Du sang artériel et veineux de huit porcs a été utilisé pour évaluer la dilution en série du sang et son effet correspondant sur l'intensité du signal de la séquence OS. In vivo: Vingt-deux volontaires en santé ont subi OS-CMR. Les concentrations d'hémoglobine (Hb) ont été mesurées au niveau de base et immédiatement après une l'infusion cristalloïde rapide de 1000 mL de solution Lactate Ringer's (LRS). Les images OS-CMR ont été prises dans une vue mid-ventriculaire court axe. L'intensité du signal myocardique a été mesurée durant une rétention respiratoire volontaire maximale, suite à une période d'hyperventilation de 60 secondes. Les changements dans l'intensité du signal entre le début et la fin de la rétention de la respiration ont été exprimés relativement au niveau de base (% de changement). Résultats: L'infusion a résulté en une diminution significative de l'Hb mesurée (142.5±3.3 vs. 128.8±3.3 g/L; p<0.001), alors que l'IS a augmenté de 3.2±1.2% entre les images du niveau de base en normo- et hypervolémie (p<0.05). L'IS d'hyperventilation ainsi que les changements d'IS induits par l'apnée ont été attenués après hémodilution (p<0.05). L'évaluation quantitative T2* a démontré une corrélation négative entre le temps de T2* et la concentration d'hémoglobine (r=-0.46, p<0.005). Conclusions: Il existe plusieurs éléments confondants de la technique OS-CMR qui requièrent de l'attention et de l'optimisation pour une future implémentation clinique à grande échelle. Le statut d'hydratation en particulier pourrait être un élément confondant dans l'imagerie OS-CMR. L'hypervolémie mène à une augmentation en IS au niveau de base et atténue la réponse IS durant des manoeuvres de respiration vasoactives. Cette atténuation de l'intensité du signal devrait être tenue en compte et corrigée dans l'évaluation clinique d'images OS-CMR.
Background: Oxygenation-sensitive cardiovascular magnetic resonance (OS-CMR) has become a feasible diagnostic imaging modality for monitoring changes of myocardial oxygenation. This technique has great potential for use as a primary diagnostic tool for cardiovascular disease, particularly non-invasive detection of ischemia. Yet, there are several potential confounding factors of this technique, some methodological, such as sequence parameters and others are physiological and not well understood. Due to T2 effects caused by tissue water content, the hydration status may impact signal intensity. This is one physiological aspect in particular that we aimed at quantifying the confounding effect by manipulating hydration status in humans and observing signal intensity (SI) changes in OS-CMR images. Methods: In vitro: Arterial and venous blood from eight swine were used to assess serial dilution of blood and it corresponding effect on OS sequence signal intensity. In vivo: Twenty-two healthy volunteers underwent OS-CMR. Hemoglobin (Hb) concentrations were measured at baseline and immediately following rapid crystalloid infusion of 1,000ml of Lactated Ringer’s solution (LRS). OS-CMR images were acquired in a mid-ventricular short axis view. Myocardial SI was measured during a maximal voluntary breath-hold, after a 60-second period of hyperventilation. SI changes were expressed relative to baseline (% change). Results: The infusion resulted in a significant decrease in measured Hb (142.5±3.3 vs. 128.8±3.3 g/L; p<0.001), while SI increased by 3.2±1.2% between baseline images at normo- and hypervolemia (p<0.05). Both hyperventilation SI and the SI changes induced by apnea were attenuated after hemodilution (p<0.05). Quantitative assessment showed a negative correlation between T2* and hemoglobin concentration (r=-0.46, p<0.005). Conclusions: There are several confounders to the OS-CMR technique that require attention and optimization for future larger scale clinical implementation. The hydration status in particular may be a confounder in OS-CMR imaging. Hypervolemia leads to an increase in SI at baseline and attenuates the SI response during vasoactive breathing maneuvers. This attenuation in signal intensity would need to be accounted for and corrected in clinical assessment of OS-CMR images.

The aim of this doctoral thesis was to study the incidence, risk factors and outcomes of type 1 diabetes for children in South Australia, born from 1999-2013. The incidence of type 1 diabetes has doubled in the last four decades in many countries including Australia, and has substantial individual and economic consequences. Evidence from studies on type 1 diabetes aetiology and its implications is mixed. In this thesis, the linkage of multiple population-wide administrative data over 15 years, and use of rigorous epidemiological approaches has resulted in a better understanding of the risk factors and implications of type 1 diabetes. There are four studies in this doctoral thesis. In the first descriptive study, the incidence of type 1 diabetes was estimated by individual and area-level socioeconomic characteristics among children (aged ≤11 years) in South Australia, born from 2002-2013. Findings of the study showed that type 1 diabetes incidence rates differed depending on the measures of socioeconomic characteristics. Individual-level indicators showed higher type 1 diabetes incidence among more advantaged children, however, there was no clear area-level socioeconomic patterning of type 1 diabetes. Area-level measures of socioeconomic position are likely to have a greater risk of misclassification from true socioeconomic position, which suggests that the use of area-level measures may be misleading. Socioeconomic position is a major determinant of health and can modify the risk factors of type 1 diabetes. For example, as per hygiene hypothesis, the socioeconomically dis-advantaged children are less likely to have type 1 diabetes, which is supported by the findings of individual-level socioeconomic patterning of type1 diabetes in the first study. In addition, socioeconomically disadvantaged women are less likely to have a caesarean birth and more likely to smoke in pregnancy. I chose to study these two risk factors of type 1 diabetes because the evidence was inconsistent, and some studies had methodical limitations. Evidence about the effect of caesarean section on childhood type 1 diabetes is mixed; ranging from very small or no risk to 20-30% increased risk. A prevailing theory is that exposure to the gut and vaginal microbiota during a vaginal birth protects against type 1 diabetes. Therefore, in the second study, the impact of caesarean birth on childhood type 1 diabetes (aged ≤15 years) was estimated. This involved linking multiple administrative datasets of children in South Australia, born from 1999-2013. The question was extended to whether type 1 diabetes risk differed for children born by prelabour or intrapartum caesarean to further test the idea of microbiota exposure on type 1 diabetes. That is because children born by prelabour caesarean do not get exposure to maternal vaginal microbiota, and intrapartum caesarean births may have some exposure. Findings of the study obtained from Cox proportional hazard regression analysis showed a negligible 5% higher incidence (HR = 1.05, 95% CI 0.86-1.28) for caesarean births compared with normal vaginal delivery, with wide confidence intervals including the null. Contrary to the hypothesis of a higher type 1 diabetes risk for prelabor caesarean (because of non-exposure to maternal vaginal microbiota) type 1 diabetes risk for intrapartum caesarean was slightly higher (HR = 1.08, 95% CI 0.82-1.41) than prelabor caesarean (HR = 1.02, 95% CI 0.79-1.32). This negligible risk of type 1 diabetes for children who had caesarean birth, either prelabor or intrapartum, and the potential for unmeasured confounding suggested that birth method induced variation in neonatal microbiota might not be involved in modifying type 1 diabetes risk. Like caesarean section, maternal smoking in pregnancy is also a debated risk factor for childhood type 1 diabetes. Evidence about maternal smoking on childhood type 1 diabetes is inconsistent; studies have been small, and many did not adjust for important confounders or address missing data. In the third study of this doctoral thesis, the effect of maternal smoking in pregnancy on childhood type 1 diabetes was estimated using Cox proportional hazard regression analysis, once again by linking multiple administrative datasets of children in South Australia, born from 1999-2013. The analytical approach for this study ranged; from Cox proportional hazard analysis with adjustment for wide range of confounders using the SA ECDP linked data, involving multiple imputation for missing data; to conducting meta-analysis in order to get more precise estimate. But smoking is notoriously residually confounded, therefore, I made special efforts to investigate the possibility of residual confounding by using a negative control and E-value. The findings demonstrated that maternal smoking in pregnancy was associated with a 16% (HR 0.84, 95% CI 0.67, 1.08) lower childhood type 1 diabetes incidence, compared with unexposed children, which was also supported by the meta-analytic estimates of population-based cohort studies (HR 0.72, 95% CI 0.62, 0.82) and case-control studies (OR 0.71, 95% CI 0.55, 0.86). The negative control outcome and E-value analyses indicated the potential for residual confounding in the effect of maternal smoking on childhood type 1 diabetes. Triangulation of evidence from this study along with the results of similar population-based studies, suggested a small reduced risk of childhood type 1 diabetes for children exposed to maternal smoking in pregnancy. However, the mechanisms linking maternal smoking in pregnancy with childhood type 1 diabetes require further investigation. In the fourth study of this thesis, the impact of childhood type 1 diabetes on children’s educational outcomes in year/grade 5 at age ~10 were estimated, linking population-wide data of children in South Australia, born from 1999-2005. In this study, a doubly-robust analytical method called augmented inverse probability weighting (AIPW) was used to compute the average treatment effect of type 1 diabetes on children’s educational outcomes. AIPW gives an unbiased estimate if either the outcome model or the treatment model is correctly specified. The findings of this study demonstrated that children with type 1 diabetes are not disadvantaged in terms of educational outcomes in year 5, potentially reflecting improvement in type 1 diabetes management in Australia. In summary, the work in this doctoral thesis has demonstrated that type 1 diabetes incidence differed depending on the measure of socioeconomic position. The hygiene hypothesis was only supported by the individual-level socioeconomic pattering of type 1 diabetes incidence in South Australia. The involvement of birth method induced variation in neonatal microbiota in type 1 diabetes was not supported by the caesarean and childhood type 1 diabetes study. Despite the evidence of residual confounding in the estimate of maternal smoking in pregnancy on childhood type 1 diabetes, triangulation of the evidence suggested small reduced risk for children exposed to maternal smoking in pregnancy, but further research will be needed to understand the mechanism. The findings of similar educational outcomes for children with and without type 1 diabetes, highlighted the importance of improvements in diabetes management.
Thesis (Ph.D.) -- University of Adelaide, School of Public Health, 2020