<div>With the rapid development in high-throughput technologies and the next generation sequencing (NGS) during the past decades, the bottleneck for advances in computational biology and bioinformatics research has shifted from data collection to data analysis. As one of the central goals in precision health, understanding and interpreting high-dimensional biomedical data is of major interest in computational biology and bioinformatics domains. Since significant effort has been committed to harnessing biomedical data for multiple analyses, this thesis is aiming for developing new machine learning approaches to help discover and interpret the complex mechanisms and interactions behind the high dimensional features in biomedical data. Moreover, this thesis also studies the prediction of post-treatment response given histopathologic images with machine learning.</div><div><br></div><div>Capturing the important features behind the biomedical data can be achieved in many ways such as network and correlation analyses, dimensionality reduction, image processing, etc. In this thesis, we accomplish the computation through co-expression analysis, survival analysis, and matrix decomposition in supervised and unsupervised learning manners. We use co-expression analysis as upfront feature engineering, implement survival regression in deep learning to predict patient survival and discover associated factors. By integrating Cox proportional hazards regression into non-negative matrix factorization algorithm, the latent clusters of human genes are uncovered. Using machine learning and automatic feature extraction workflow, we extract thirty-six image features from histopathologic images, and use them to predict post-treatment response. In addition, a web portal written by R language is built in order to bring convenience to future biomedical studies and analyses.</div><div><br></div><div>In conclusion, driven by machine learning algorithms, this thesis focuses on the integrative analysis given multimodal biomedical data, especially the supervised cancer patient survival prognosis, the recognition of latent gene clusters, and the application of predicting post-treatment response from histopathologic images. The proposed computational algorithms present its superiority comparing to other state-of-the-art models, provide new insights toward the biomedical and cancer studies in the future.</div>

This study investigated international medical graduates (IMGs), registered between January 1, 2007 and April 14, 2011, at the Access Centre for Internationally Educated Health Professionals in Ontario. By way of logistic regression in a cross-sectional design, it was found that permanent residents who were recent immigrants had lesser chances of being employed full-time at registration (baseline). By way of survival analysis in a cohort design, it was found that younger IMGs who have been in Canada less than 5 years and who have taken the Medical Council of Canada Evaluating Exam (MCCEE) have the greatest chances of securing residency positions in Canada or the US, whereas IMGs from Eastern Europe, South Asia and Africa have lesser chances. It was revealed that registered IMGs are a vulnerable population, and certain groups may be disadvantaged due to underlying characteristics. These groups can be targeted for specific interventions.

<p> The sales of deer licenses, one of the most important revenue sources for wildlife management at the Indiana Department of Natural Resources (IDNR), have been declining for a decade. To increase its funds, the agency is considering launching a new lifetime deer license, which would allow hunters to harvest deer (and possibly other species) each year for the rest of their lives in exchange for a large, up-front fee. The forward-looking nature of the decision to buy a lifetime license means hunters’ choice behavior is necessarily dynamic. We estimate a dynamic discrete choice model using data from a discrete choice experiment (DCE) to capture this forward-looking choice behavior and to estimate hunters’ preferences for different lifetime license designs. We find that our dynamic model better fits our data than a standard, static choice model. We also find that hunters prefer licenses that allow (i) harvest of antlered and antlerless deer to one that only allows harvest of antlerless deer and (ii) harvest of additional species beyond just deer. We use our model to estimate the price of lifetime licenses that maximizes IDNR revenues. This is the first study to estimate the value of lifetime deer hunting licenses using a dynamic approach. This dynamic approach can help improve the IDNR’s decision-making to maximize its revenue and stabilize wildlife management funds.  </p>

La consanguinité, soit l'union productive de conjoints partageant des allèles identiques provenant d'un ancêtre commun, s'est accumulée au fil du temps au Québec ancien. Parallèlement, le Québec a été victime de plusieurs épidémies. Le but de cette étude est d'évaluer la relation entre la mortalité des enfants et la consanguinité dans les périodes épidémiques du Québec ancien entre 1720 et 1830. D'une part, l'hypothèse émise est que les enfants ayant des gènes homologues sur plusieurs loci auraient un taux de mortalité significativement plus élevé comparativement aux enfants non consanguins, en raison du désavantage des homozygotes. D'autre part, les individus consanguins peuvent avoir une survie plus favorable en raison de l'effet d’enracinement, combien de générations une famille est établie dans la colonie, présent dans la mesure de la consanguinité. De plus, l'avantage social d'une famille étroitement liée peut favoriser la survie de l'enfant en accordant plus de soutien social aux parents et de surveillance de l'enfant. Les courbes de survie de Kaplan-Meier sont représentées graphiquement et des modèles de régression de Cox sont exécutés pour explorer et démêler partiellement les rôles des facteurs génétiques et environnementaux. Les immigrants, les naissances multiples et les individus sans généalogie du Registre de la population du Québec ancien (RPQA) et de l'Infrastructure intégrée des microdonnées historiques de la Population du Québec (IMPQ) sont exclus. Au total, 610 412 individus sont analysés dans les modèles de Cox. Les rapports de risque pour les épidémies augmentent avec l'âge et les rapports de risque pour la consanguinité éloignée ressemblent souvent au groupe référence, les non consanguins. De plus, les effets diffèrent selon le sexe et le groupe d'âge. Généralement, si les enfants avec une consanguinité proche, ceux identifiés comme consanguins avec seulement trois générations ascendantes, ne subissent pas de surmortalité dans un groupe d'âge précédent, les modèles de Cox signalent une survie défavorable de ces individus lors des épidémies. Des effets sous-jacents tels que des processus de sélection et des variables de contrôle relatives à l’enracinement peu robustes guident les résultats de l'interaction entre les épidémies et la consanguinité, de sorte que la prémisse reste à valider. / Consanguinity, the productive union of spouses sharing identical alleles from a common ancestor, accumulated over time in Colonial Quebec. Concurrently, Quebec was the victim of several epidemics. The aim of this study is to evaluate the relationship between child mortality and consanguinity in epidemic periods of Colonial Quebec between 1720 and 1830. On the one hand, it is hypothesized that children with homologous genes on many loci would have a significantly higher mortality rate compared to non consanguineous children, due to homozygote disadvantage. On the other hand, consanguineous individuals may have a more favourable survival because of the effect of settlement, how many generations a family has been in the colony, present in the measure of consanguinity. Further, the social benefit of a closely bound family may favour child survival by providing more social support to the parents and child supervision. Kaplan-Meier survival curves are graphed, and Cox regression models are run to explore and partially disentangle the roles of genetic and environmental factors. Immigrants, multiple births and individuals lacking a genealogy from the Registre de population du Québec ancien (RPQA) and Infrastructure intégrée des microdonnées historiques de la Population du Québec (IMPQ) are excluded. Altogether, 610,412 individuals are analysed in the Cox models. Hazard ratios for epidemics increase with age and distant consanguinity hazard ratios often resemble the no consanguinity reference group. Further, the effects differ by sex and age group. Generally, if closely consanguineous children, those identified as consanguineous with only three ascending generations, do not undergo excess mortality in a previous age group, the Cox models signal an unfavourable survival of these individuals during epidemics. Underlying effects such as selection processes and unrobust control variables for settlement guide the results of the interaction between epidemics and consanguinity, so the premise, though convincing, remains to be validated.

In recent years, the transplant community has explored and adopted tools for quantifying clinical insight into illness severity and frailty. This dissertation work explores the interplay between objective and subjective assessments of physical health status and the implications for liver transplant candidate and recipient outcomes. The first aim characterizes national epidemiologic trends and the impact of Centers for Medicare and Medicaid quality improvement policies on likelihood of waitlist removal based on the patient being too frail to benefit from liver transplant (“too sick to transplant”). This aim includes more than a decade (2002–2012) of comprehensive national transplant waitlist data (Scientific Registry of Transplant Recipients (SRTR)). The second aim will assess and define objective parameters of liver transplant patient frailty by measuring decline in lean core muscle mass (“sarcopenia”) using abdominal CT scans collected retrospectively at a single U.S. transplant center between 2006 and 2015. The relationship between these objective sarcopenia measures and subjective functional status assessed using the Karnofsky Functional Performance (KPS) scale are described and quantified. The third aim quantifies the extent to which poor functional status (KPS) pre-transplant is associated with worse post-transplant survival and includes national data on liver transplantations conducted between 2005 and 2014 (SRTR). The results of this dissertation will help providers in the assessment of frailty and subsequent risk of adverse outcomes and has implications for strategic clinical management in anticipation of surgery. This research will also to serve to inform national policy on the design of transplant center performance measures.

Background: Overtreatment of localized prostate cancer (PCa) is a concern as many men die of other causes prior to experiencing a treatment benefit. This dissertation characterizes the need for assessing other cause mortality (OCM) risk in older men with PCa and informs efforts to identify patients most likely to benefit from definitive PCa treatment. Methods: Using the linked Surveillance Epidemiology and End Results-Medicare Health Outcomes Survey database, 2,931 men (mean age=75) newly diagnosed with clinical stage T1a-T3a PCa from 1998-2009 were identified. Survival analysis methods were used to compare observed 10-year OCM by primary treatment type. Age and health factors predictive of primary treatment type were assessed with multinomial logistic regression. Predicted mortality estimates from Social Security life tables (recommended for life expectancy evaluation) and two OCM risk estimation tools were compared to observed rates. An improved OCM prediction model was developed fitting Fine and Gray competing risks models for 10-year OCM with age, sociodemographic, comorbidity, activities of daily living, and patient-reported health data as predictors. The tools’ ability to discriminate between patients who died and those who did not was evaluated with Harrell’s c-index (range 0.5-1), which also guided new model selection. Results: Fifty-four percent of older men with localized PCa underwent radiotherapy while 13% underwent prostatectomy. Twenty-three percent of those treated with radiotherapy and 12% of those undergoing prostatectomy experienced OCM within 10 years of treatment and thus were considered overtreated. Health factors indicative of a shorter life expectancy (increased comorbidity, worse physical health, smoking) had little to no association with radiotherapy assignment but were significantly related to reductions in the likelihood of undergoing prostatectomy. Social Security life tables overestimated mortality risk and discriminated poorly between men who died and those who did not over 10 years (c-index=0.59). Existing OCM risk estimation tools were less likely to overestimate OCM rates and had limited but improved discrimination (c-index=0.64). A risk model developed with self-reported age, Charlson comorbidity index score, overall health (excellent-good/fair/poor), smoking, and marital status predictors had improved discrimination (c-index=0.70). Conclusions: Overtreatment of older men with PCa is primarily attributable to radiotherapy and may be reduced by pretreatment assessment of mortality-related health factors. This dissertation provides a prognostic model which utilizes a set of five self-reported characteristics that better identify patients likely to die of OCM within 10 years of diagnosis than age and comorbidity-based assessments alone.

<p>UVOD: Akutna mijeloidna leukemija (AML) predstavlja heterogenu grupu oboljenja u odnosu na morfologiju, citogenetiku, molekularnu genetiku, zbog čega se deli na različite kliničke i biolo&scaron;ke entitete, sa različitim odgovorom na terapiju i ishodom lečenja. Humani EVI1 (ecotropic virus integration-1) gen ima ulogu multifunkcionalnog nuklearnog transkripcionog faktora, kako u normalnoj tako i u malignoj hematopoezi. Sve je vi&scaron;e istraživanja koja ističu negativni prognostički značaj visoke ekspresije (overexpression) EVI1 gena u AML.&nbsp; CILJEVI: Ciljevi ovog istraživanja su da se ispita klinički i prognostički značaj ekspresije gena EVI1 u AML, kao i da se utvrdi povezanost visoke ekspresije gena EVI1 sa nalazima citogenetskog ispitivanja i molekularnim markerima: FLT3 mutacijom i nukleofozmin 1 (NPM1) mutacijom. MATERIJAL I METODE: Ovim prospektivnim istraživanjem je obuhvaćena grupa od 38 odraslih novodijagnostikovanih bolesnika sa de novo, non M3 AML, kod kojih je započeto standardno lečenje, a koji su dijagnostikovani i lečeni u Klinici za hematologiju Kliničkog centra Vojvodine u periodu od jula 2012. do marta 2014. Određivanje ekspresije gena EVI1 je vr&scaron;eno pomoću real time kvantitativne PCR (qPCR) metode, tehnikom TaqMan, a relativna ekspresija EVI1 gena je određena primenom &Delta;&Delta;Ct metode.&nbsp; REZULTATI: Medijana starosti bolesnika pri postavljanju dijagnoze AML je bila 52 godine (23-80). Ustanovljena je statistički značajna razlika između ekspresije gena EVI1 kod zdravih osoba (kontrolna grupa) i obolelih od akutne mijeloidne leukemije (p=0.008). Računajući relativnu ekspresiju, 13,2 % bolesnika je imalo visoku ekspresiju (overexpression) gena EVI1. U odnosu na kliničke i laboratorijske karakteristike bolesnika (kao &scaron;to su pol, starost, parametri krvne slike, nivo laktat dehidrogenaze, procenat blasta u perifernoj krvi i ko&scaron;tanoj srži, potom tip akutne mijeloidne leukemije, performans status, komorbiditetni indeks) nije ustanovljena statistički značajna razlika između bolesnika sa visokom ekspresijom EVI1 gena i ostalih bolesnika. Postoji statistički značajna povezanost visoke ekspresije EVI1 gena i nepostojanja NPM1 mutacije (p=0,031), kao i između visoke ekspresije EVI1 gena i prisustva monozomije 7 (p=0,047). Visoka ekspresija EVI1 gena je povezana sa kraćim preživaljvanjem bez dogaĎaja (p=0,004), kao i sa kraćim ukupnim preživljavanjem (p=0,025).&nbsp; ZAKLJUČCI: Postoji značajno povećana ekspresija gena EVI1 kod obolelih od AML u odnosu na zdrave kontrole. Visoka ekspresija EVI1 gena je faktor lo&scaron;e prognoze kod obolelih od akutne mijeloidne leukemije i u kombinaciji sa drugim prognostičkim markerima, doprinosi boljoj risk stratifikaciji ovih bolesnika.</p> / <p>INTRODUCTION: Acute myeloid leukaemia (AML) represents a heterogenous group of diseases in terms of morphology, cytogenetics, molecular genetics, so it can be divided into distinct clinical and biological entities, with variable responsiveness to therapy and different treatment outcome. Human EVI1 (ecotropic virus integration-1) gene plays a role of multifunctional nuclear transcriptional factor, not only in normal, but also in malignant haematopoiesis. There are more and more investigations indicating high EVI1 expression (EVI1 overexpression) as a negative prognostic marker in AML.&nbsp; PURPOSES: The main goal of this investigation was to examine the clinical and prognostic significance of EVI1 expression in AML, as well as to investigate whether there was any association of EVI1 overexpression with cytogenetic abnormalities and other standard molecular prognostic factors, such as FLT3 mutation and nucleophosmin 1 (NPM1) mutation.&nbsp; PATIENTS AND METHODS: This prospective study included 38 adult newly diagnosed patients with de novo nonM3 AML, in whom a standard treatment was started at Clinic of Haematology, Clinical center of Vojvodina in the period from July 2012 to March 2014. EVI1 expression was analyzed by real-time quantitative polymerase chain reaction using TaqMan, and relative EVI1 expression was determined by &Delta;&Delta;Ct method.&nbsp; RESULTS: Median age of patients at diagnosis was 52 (aged 23-80). There has been determined statistically higher EVI1 expression in our AML patients than in healthy volunteers (control group) (p=0.008). The relative EVI1 overexpression was observed in 13.2% of the patients. No significant differences in clinical and laboratory patient data (including sex, age, whole blood counts, lactate dehydrogenase level, peripheral and bone marrow blast percentages, type of AML, performance status, comorbidity index) were observed between patients with high EVI1 expression and patients without high EVI1 expression. Our investigation revealed inverse correlation of high EVI1 expression and nucleophosmin 1 mutation (p=0,031). Also high EVI1 expression was significantly associated with monosomy 7 (p=0,047). Survival analysis revealed significantly inferior event free survival (p=0,004) and overall survival (p=0,025) for patients with high EVI1 expression compared to the other patients.&nbsp; CONCLUSION: EVI1 expression is significantly higher in AML patients compared to healthy controls. High EVI1 expression is a poor prognostic marker for patients with AML, and in combination with other well established prognostic markers, contributes to better risk stratification of these patients.</p>

MSc (Statistics) / Department of Statistics / Selection of variables is vital in high dimensional statistical modelling as it aims to identify the right subset model. However, variable selection for discrete survival analysis poses many challenges due to a complicated data structure. Survival data might have unobserved heterogeneity leading to biased estimates when not taken into account. Conventional variable selection methods have stability problems. A simulation approach was used to assess and compare the performance of Least Absolute Shrinkage and Selection Operator (Lasso) and gradient boosting on discrete survival data. Parameter related mean squared errors (MSEs) and false positive rates suggest Lasso performs better than gradient boosting. Frailty models outperform discrete survival models that do not account for unobserved heterogeneity. The two methods were also applied on Zimbabwe Demographic Health Survey (ZDHS) 2016 data on age at first marriage and did not select exactly the same variables. Gradient boosting retained more variables into the model. Place of residence, highest educational level attained and age cohort are the major influential factors of age at first marriage in Zimbabwe based on Lasso. / NRF

<p>Karcinom bronha je danas u svetu najče&scaron;ći uzrok smrti povezanih sa malignim bolestima. U XX veku je registrovan značajan porast kako incidence, tako i mortaliteta karcinoma bronha u većini zemalja. Medijana&nbsp; preživljavanja u svim stadijumima bolesti se značajno pobolj&scaron;ala poslednjih godina XX veka, ali nedovoljno u odnosu na očekivano. U najvećem broju slučajeva, bolest se otkriva u uznapredovalom stadijumu, kada je radikalno hirur&scaron;ko lečenje kao optimalan vid lečenja nemoguće. Određivanje stadijuma bolesti (stejdžing) je najbitniji segment u evaluaciji svakog bolesnika s karcinomom bronha. Utvrđivanje zahvaćenosti medijastinalnih&nbsp; limfnih čvorova karcinomom je od posebne važnosti, jer je u velikom broju slučajeva upravo nodalni status faktor koji određuje svsishodnost primene hirur&scaron;kog lečenja, radioterapije i hemioterapije, a samim tim i jedan od bitnih faktora prognoze bolesnika sa nemikrocelularnim karcinomom bronha NSCLC. Bolesnici sa dokazanom zahvaćeno&scaron;ću N2 medijastinalnih limfnih čvorova se svrstavaju u IIIA stadijum NSCLC koji je potencijalno resektabilan, dok se bolesnici sa dokazanom zahvaćeno&scaron;ću N3 medijastinalnih limfnih čvorova svrstavaju u IIIB stadijum NSCLC, koji se smatra neresektabilnim. Cilj ove doktorske disertacije je bio da se utvrdi da li postoje prognostički značajni faktori za rezultat lečenja medijastinoskopski dokazanog N2 i N3 stadijuma NSCLC. Studija je bila nerandomizovana, delom retrospektivnog, a delom prospektivnog karaktera. U ispitivanje je uključeno 60 bolesnika lečenih u Institutu za plućne bolesti Vojvodine tokom&nbsp; 2006., 2007. i 2008. godine. Kod svih uključenih bolesnika medijastinoskopijom je dokazana propagacija NSCLC u medijastinalne limfne čvorove. U radu su analizirani sledeći faktori: pol, starost, ECOG performans status, pridružena hronična opstruktivna bolest pluća (HOBP), pridruženo kardiovaskularno oboljenje sa simtomatologijom klasifikovanom prema NYHA, T faktor, lokalizacija i broj medijastinoskopski dokazanih metastatski zahvaćenih limfnih čvorova, vrsta primenjenog&nbsp; lečenja (hemioradioterapija, hemioterapija, operacija), rezultat lečenja (odgovor na terapiju i preživljavanje). Univarijantnom analizom je utvrđeno da su kod bolesnika sa medijastinoskopski dokazanim N2 i N3 stadijumom NSCLC prognostički faktori koji su imali uticaj na lo&scaron;ije preživljavanje bili: ECOG PS 2 (p=0,00000), pridruženo kardivaskularno oboljenja sa simptomatologijom klase NYHA II (p=0,00113), zahvaćenost kontralateralnih medijastinalnih medijastinalnih limfnih čvorova (N3 stadijum) (p=0,000003), dok je uticaj zahvaćenosti vi&scaron;e pozicija ipsilateralnih medijastinalnih limfnih čvorova (multi station N2) bio na granici statističke značajnosti (p=0,05385). Utvrđeno je da bolesnici sa N2 i N3 stadijumom NSCLC lečeni hemioradioterapijom imaju bolju stopu odgovora na primenjenu terapiju u odnosu na bolesnike lečene samo hemioterapijom (p=0,03118), kao i da operativno lečenje primenjeno kod bolesnika koji su imali dobar odgovor na sprovedenu terapiju ima statistički značajan uticaj u vidu boljeg preživljavanja (p=0,00121). Univarijantnom analizom nije utvrđen značajan uticaj sledećih faktora na preživljavanje bolesnika sa N2 i N3 stadijomom NSCLC: pol, starost, pridružena HOBP, skvamozni tip NSCLC i T faktor. Multivarijantnom analizom su kao nezavisni prognostički faktori na preživljavanje bolesnika sa N2 i N3 stadijumom NSCLC utvrđeni klinički N status (bolje preživljavanje ima N2 u odnosu na N3 stadijum) i sprovedena terapija (bolje preživljavanje ima hemioradioterapija u odnosu na hemioterapiju). Dobijeni rezultati navode nas na zaključak da su pozicija i broj zahvaćenih pozicija medijastinalnih limfnih čvorova, koji su utvrđeni medijastinoskopski, kao i sprovođenje multimodalnog lečenja ključni prognostički faktori za preživljavanje bolesnika sa N2 i N3 stadijumom NSCLC.</p> / <p>Lung&nbsp; cancer&nbsp; is&nbsp; the&nbsp; most&nbsp; common cause of cancer&nbsp; related mortality&nbsp; worldwide.&nbsp; Increase&nbsp; in&nbsp; both&nbsp; incidence&nbsp; and mortality&nbsp; of&nbsp; lung&nbsp; cancer&nbsp; was&nbsp; registered&nbsp; throughout&nbsp; 20th century. The median survival in every stage of lung cancer has been improved in last years of 20th century but it is still not satisfactory. In most cases, lung cancer is diagnosed in advanced&nbsp; stage&nbsp; when&nbsp; surgical&nbsp; treatment&nbsp; as&nbsp; the&nbsp; optimal approach&nbsp; is&nbsp; not&nbsp; possible. Staging&nbsp; is&nbsp; the&nbsp; most&nbsp; important element&nbsp; in&nbsp; the&nbsp; evaluation&nbsp; of&nbsp; every&nbsp; lung&nbsp; cancer&nbsp; patient. Mediastinal lymph node involvement is crucial, because in most of the cases nodal staging is factor which determines appropriate use of surgery, radiotherapy and chemotherapy and it is one of the important factors influencing prognosis of lung cancer patients. Patients with proven involvement of ipsilateral&nbsp; mediastinal&nbsp; lymph&nbsp; nodes&nbsp; (N2&nbsp; stage)&nbsp; are categorized&nbsp; in&nbsp; IIIA&nbsp; stage&nbsp; which&nbsp; is&nbsp; considered&nbsp; to&nbsp; be potentially resectable, and patients with proven involvement of&nbsp; contralateral&nbsp; mediastinal&nbsp; lymph&nbsp; nodes&nbsp; (N3&nbsp; stage)&nbsp; are categorized&nbsp; in&nbsp; IIIB&nbsp; stage,&nbsp; which&nbsp; is&nbsp; considered&nbsp; to&nbsp; be nonresectable. The aim of this study was the determination of&nbsp; significant&nbsp; prognostic&nbsp; factors&nbsp; that&nbsp; have&nbsp; influence&nbsp; on treatment&nbsp; and&nbsp; survival&nbsp; of&nbsp; non-small&nbsp; cell&nbsp; lung&nbsp; cancer (NSCLC)&nbsp; patients&nbsp; in&nbsp; stage&nbsp; N2&nbsp; and&nbsp; N3.&nbsp; Study&nbsp; was nonrandomized,&nbsp;&nbsp; partially&nbsp;&nbsp; retrospective&nbsp;&nbsp; and&nbsp;&nbsp; partially prospective. It included 60 patients treated at the Institute for Pulmonary&nbsp; Diseases&nbsp; of&nbsp; Vojvodina&nbsp; during&nbsp; 2006,&nbsp; 2007&nbsp; and 2008. Cancer involvement of mediastinal lymph nodes was determined by mediastinoscopy in every patient. In study we analyzed following factors: gender, age, ECOG performance&nbsp;&nbsp; status,&nbsp;&nbsp; associated&nbsp;&nbsp; chronic&nbsp;&nbsp; obstructive pulmonary&nbsp; disease&nbsp; (COPD),&nbsp; associated&nbsp; cardiovascular disease with symptoms graded by NYHA classification, T status, position and number of involved mediastinal lymph nodes,&nbsp; applied&nbsp; treatment&nbsp; (surgery,&nbsp; chemoradiotherapy, chemotherapy alone), treatment result (response to treatment and&nbsp; survival).&nbsp; Prognostic&nbsp; factors&nbsp; for&nbsp; poorer&nbsp; survival&nbsp; on univariant analysis were ECOG PS 2 (p=0,0000), associated cardiovascular&nbsp;&nbsp; disease&nbsp;&nbsp; with&nbsp;&nbsp; symptoms&nbsp;&nbsp; NYHA&nbsp;&nbsp; II (p=0,00113)&nbsp; and&nbsp; involvement of&nbsp; contralateral&nbsp; mediastinal lymph nodes (N3 stage) (p=0,00003) while multi station N2 disease was borderline significant at level of p=0,05385. It was determined that patients treated with chemoradiotherapy achieved better response to treatment compared to patients treated&nbsp; with&nbsp; chemotherapy&nbsp; alone&nbsp; (p=0,03118).&nbsp; Univariant analyses&nbsp; did&nbsp; not&nbsp; confirm&nbsp; significance&nbsp; of&nbsp; gender,&nbsp; age, associate COPD, squamous cell lung cancer and T factor on survival. Multivariante&nbsp; analyses&nbsp; identified&nbsp; N&nbsp; status&nbsp; (better survival has N2 stage compared to N3 stage of NSCLC) and conducted treatment (better survival has&nbsp; chemoradiotherapy compared to chemotherapy alone) as independent prognostic factors.&nbsp; Our&nbsp; results&nbsp; suggest&nbsp; that&nbsp; position&nbsp; and&nbsp; number&nbsp; of cancer involved&nbsp; mediastinal lymph nodes position,&nbsp; proven by mediastinoscopy, as well as the conducted multimodality treatment are key prognostic factors which might influence the survival of patients with N2 and N3 stage of NSCLC.</p>

Dans ce travail, nous nous intéressons à l'inférence statistique dans deux modèles semi-paramétrique et non-paramétrique stratifiés de durées de vie censurées. Nous proposons tout d'abord une statistique de test d'ajustement pour le modèle de régression stratifié à risques proportionnels. Nous établissons sa distribution asymptotique sous l'hypothèse nulle d'un ajustement correct du modèle aux données. Nous étudions les propriétés numériques de ce test (niveau, puissance sous différentes alternatives) au moyen de simulations. Nous proposons ensuite une procédure permettant de stratifier le modèle à 1isques proportionnels suivant un seuil inconnu d'une variable de stratification. Cette procédure repose sur l'utilisation du test d'ajustement proposé précédemment. Une étude de simulation exhaustive est conduite pour évaluer les pe1fonnances de cette procédure. Dans une seconde partie de notre travail, nous nous intéressons à l'application du test du logrank stratifié dans un contexte de données manquantes (nous considérons la situation où les strates ne peuvent être observées chez tous les individus de l'échantillon). Nous construisons une version pondérée du logrank stratifié adaptée à ce problème. Nous en établissons la loi limite sous l'hypothèse nulle d'égalité des fonctions de risque dans les différents groupes. Les propriétés de cette nouvelle statistique de test sont évaluée au moyen de simulations. Le test est ensuite appliqué à un jeu de données médicales. / In this work, we are interested in the statistical inference in two semi-parametric and non-parametric stratified models for censored data. We first propose a goodnessof- fit test statistic for the stratified proportional hazards regression model. We establish its asymptotic distribution under the null hypothesis of a correct fit of the model. We investigate the numerical properties of this test (level, power under different alternatives) by means of simulations. Then, we propose a procedure allowing to stratify the proportional hazards model according to an unknown threshold in a stratification variable. This procedure is based on the goodness-of-fit test proposed earlier. An exhaustive simulation study is conducted to evaluate the performance of this procedure. In a second part of our work, we consider the stratified logrank test in a context of missing data (we consider the situation where strata can not be observed on all sample individuals). We construct a weighted version of the stratified logrank, adapted to this problem. We establish its asymptotic distribution under the null hypothesis of equality of the hazards functions in the different groups. The prope1ties of this new test statistic are assessed using simulatious. Finally, the test is applied to a medical dataset.