2,781 research outputs found
Appendix -Supplemental material for Adjustment for time-dependent unmeasured confounders in marginal structural Cox models using validation sample data
No full textSupplemental material, Appendix for Adjustment for time-dependent unmeasured confounders in marginal structural Cox models using validation sample data by Rebecca M Burne and Michal Abrahamowicz in Statistical Methods in Medical Research</p
Self-rated health in Canadian immigrants:analysis of the Longitudinal Survey of Immigrants to Canada
No full textAbstract not availableManinder Singh Setia, John Lynch, Michal Abrahamowicz, Pierre Tousignant, Amelie Quesnel-Valle
SMM902179 Supplemental Material - Supplemental material for Modeling of cumulative effects of time-varying drug exposures on within-subject changes in a continuous outcome
No full textSupplemental material, SMM902179 Supplemental Material for Modeling of cumulative effects of time-varying drug exposures on within-subject changes in a continuous outcome by Coraline Danieli, Therese Sheppard, Ruth Costello, William G Dixon and Michal Abrahamowicz in Statistical Methods in Medical Research</p
Abrahamowicz M: Using generalized additive models to reduce residual confounding. Stat Med 2004
No full textSUMMARY Traditionally, confounding by continuous variables is controlled by including a linear or categorical term in a regression model. Residual confounding occurs when the e ect of the confounder on the outcome is mis-modelled. A continuous representation of a covariate was previously shown to result in a less biased estimate of the adjusted exposure e ect than categorization provided the functional form of the covariate-outcome relationship is correctly speciÿed. However, this is rarely known. In contrast to parametric regression, generalized additive models (GAM) ÿt a smooth dose-response curve to the data, without requiring a priori knowledge of the functional form. We used simulations to compare parametric multiple logistic regression vs its non-parametric GAM extension in their ability to control for a continuous confounder. We also investigated several issues related to the implementation of GAM in this context, including: (i) selecting the degrees of freedom; and (ii) alternative criteria for inclusion=exclusion of the potential confounder and for choosing between parametric and non-parametric representation of its e ect. The impact of the shape and strength of the confounder-disease association, sample size, and the correlation between the confounder and exposure were investigated. Simulations showed that when the confounder has a non-linear association with the outcome, compared to a parametric representation, GAM modelling (i) reduced the mean squared error for the adjusted exposure e ect; (ii) avoided in ation of the type I error for testing the exposure e ect. When the true confounder-outcome relationship was linear, GAM performed as well as the parametric logistic regression. When modelling a continuous exposure non-parametrically, in the presence of a continuous confounder, our results suggest that assuming a linear e ect of the confounder and focussing on the non-linearity of the exposure-outcome relationship leads to spurious ÿndings of non-linearity: joint non-linear modelling is necessary. Overall, our results suggest that the use of GAM to reduce residual confounding o ers several improvements over conventional parametric modelling
Childhood body fatness and the risk of ovarian cancer
No full textCase-control study; Authors: Kevin L’Espérance, Michal Abrahamowicz, Jennifer O’Loughlin, Anita Koushi
Childhood body fatness and the risk of ovarian cancer
No full textCase-control study; Authors: Kevin L’Espérance, Michal Abrahamowicz, Jennifer O’Loughlin, Anita Koushi
Business plan for a person operating under a business name Individual Football Training Litol - Michal Zoubek
No full textBakalářská práce se zaměřuje na sestavení podnikatelského plánu pro potenciálně nově zakládanou firmu Individuální fotbalové tréninky Litol – Michal Zoubek, která by nabízela individuální fotbalové tréninky v Litoli na hřišti. Cílem práce je provést potřebné analýzy pro přípravu podnikatelského plánu a zhotovení jeho návrhu. V závěru autor odpoví na výzkumnou otázku, zda se v tomto konkrétním prostředí a trhu za konkrétních podmínek vyplatí začít podnikat. K dosažení cílů autor provede analýzy související s tématem, provede dotazníkové šetření, sestaví vlastní podnikatelský plán a závěrem odpoví na výzkumnou otázku.The bachelor thesis focuses on the preparation of a business plan for a potential newly established company Individual football training Litol – Michal Zoubek, which would offer individual football training in Litol. The aim of the thesis is to carry out the necessary analyses for the preparation of the business plan and to make a proposal. Finally, the author answers the research question whether it is worth starting a business in this particular environment and market under specific conditions. In order to achieve the objectives, the author will carry out analyses related to the topic, conduct a questionnaire survey, draw up his own business plan and finally answer the research question
Novel glucose lowering agents are associated with a lower risk of cardiovascular and adverse events in type 2 diabetes: A population based analysis
No full textBackground: Recent randomized control trials have described a protective cardiovascular effect of novel glucose
lowering drugs in patients at high cardiovascular risk. Whether these second-line agents have similar effects in
the general population is unknown. We aimed to compare the risk of major cardiovascular and adverse events
in new users of sodium-glucose cotransporter-2 inhibitors (SGLT-2i), dipeptidyl peptidase-4 inhibitor (DPP-
4i), glucagon-like peptide 1 agonist (GLP-1a), and sulfonylurea in T2DM patients not controlled on metformin
therapy.
Methods: Retrospective cohort study using the MarketScan database (2011–2015). We selected T2DM individuals
who were newly dispensed sulfonylureas, SGLT-2i, DPP-4i, or GLP-1a, as second-line therapy, added to metformin.
Cohort entry was defined by the date of the first prescription of the second-line agent. Time to first non-fatal
cardiovascular or adverse event was compared using Cox regression models adjusted for confounders.
Results: Among 118,341 T2DM patients using metformin (mean age: 56), most were at low cardiovascular risk
(4% with a previous cardiovascular or cerebrovascular event). During a median follow-up of 10 months compared
with sulfonylureas users, cardiovascular risk was lower in users of SGLT-2i (aHR=0.61; 95% CI: 0.40–0.97), DPP-
4i (aHR = 0.79; 95% CI: 0.69–0.90) and GLP-1a (aHR = 0.65; 95% CI: 0.48–0.89). Serious adverse events were
rare but compared with sulfonylurea, the risk was lower in new users of novel glucose-lowering agents.
Conclusion: In our analyses, which included patients with and without prior cardiovascular disease, initiating
novel glucose-lowering drugs as second-line therapy for T2DM was associated with a lower risk of cardiovascular
and adverse events than sulfonylurea initiatio
Sex Differences in Cardiovascular Effectiveness of Newer Glucose‐Lowering Drugs Added to Metformin in Type 2 Diabetes Mellitus
No full textBackgroundRandomized controlled trials showed that newer glucose-lowering agents are cardioprotective, but most participants were men. It is unknown whether benefits are similar in women.Methods and ResultsAmong adults with type 2 diabetes mellitus not controlled with metformin with no prior use of insulin, we assessed for sex differences in the cardiovascular effectiveness and safety of sodium-glucose-like transport-2 inhibitors (SGLT-2i), glucagon-like peptide-1 receptor agonists (GLP-1RA), dipeptidyl peptidase-4 inhibitors, initiated as second-line agents relative to sulfonylureas (reference-group). We studied type 2 diabetes mellitus American adults with newly dispensed sulfonylureas, SGLT-2i, GLP-1RA, or dipeptidyl peptidase-4 inhibitors (Marketscan-Database: 2011-2017). We used multivariable Cox proportional hazards models with time-varying exposure to compare time to first nonfatal cardiovascular event (myocardial infarction/unstable angina, stroke, and heart failure), and safety outcomes between drugs users, and tested for sex-drug interactions. Among 167 254 type 2 diabetes mellitus metformin users (46% women, median age 59 years, at low cardiovascular risk), during a median 4.5-year follow-up, cardiovascular events incidence was lower in women than men (14.7 versus 16.7 per 1000-person-year). Compared with sulfonylureas, hazard ratios (HRs) for cardiovascular events were lower with GLP-1RA (adjusted HR-women: 0.57, 95% CI: 0.48-0.68; aHR-men: 0.82, 0.71-0.95), dipeptidyl peptidase-4 inhibitors (aHR-women: 0.83, 0.77-0.89; aHR-men: 0.85, 0.79-0.91) and SGLT-2i (aHR-women: 0.58, 0.46-0.74; aHR-men: 0.69, 0.57-0.83). A sex-by-drug interaction was statistically significant only for GLP-1RA (P=0.002), suggesting greater cardiovascular effectiveness in women. Compared with sulfonylureas, risks of adverse events were similarly lower in both sexes for GLP-1RA (aHR-women: 0.81, 0.73-0.89; aHR-men: 0.80, 0.71-0.89), dipeptidyl peptidase-4 inhibitors (aHR-women: 0.82, 0.78-0.87; aHR-men: 0.83, 0.78-0.87) and SGLT-2i (aHR-women: 0.68, 0.59-0.78; aHR-men: 0.67, 0.59-0.78) (all sex-drug interactions for adverse events P>0.05).ConclusionsNewer glucose-lowering drugs were associated with lower risk of cardiovascular events than sulfonylureas, with greater effectiveness of GLP-1RA in women than men. Overall, they appeared safe, with a better safety profile for SGLT-2i than for GLP-1RA regardless of sex
Flexible modeling of continuous time-varying covariates in time-to-event analysis
No full textAs repeated measurements of many prognostics factors become more easily obtainable through electronic data capture, an increasing number of modern clinical and epidemiological studies start to incorporate them into the analyses. In the context of time-to-event outcomes, the repeatedly measured risk/prognostic factors are usually incorporated as time-varying covariates (TVCs) in the Cox proportional hazards (PH) model. However, adequate modeling of continuous TVCs often poses challenges because of the uncertainty about (i) the nature of the dependence of the current hazard on the current and/or past values of TVCs, (ii) the functional form of the effects of continuous TVCs on the log hazard, and (iii) potential time-dependent changes in the strength of the effect. Furthermore, in most applications, it is implicitly assumed that the values of the TVCs of interest are available at any time during the follow-up. In contrast, in practice, the available TVC measurements are usually sparse in time, which brings additional analytical challenges. The overall objective of this thesis is to enhance the modeling of the time-varying covariates in time-to-event analysis by addressing each of the aforementioned challenges. The original contributions of the thesis consist of three manuscripts, each of which proposes new methods, evaluates them in simulations, and illustrates their real-life applications. The first manuscript focuses on a relatively simpler situation, when the current hazard is assumed to be associated with only the current values of the TVCs. The conventional PH model assumes that covariate effects on the hazard are constant-over-time, and that the effects of continuous covariates on the logarithm of the hazard are linear, which may often not be the case in practice. To avoid these restrictive assumptions, we propose a flexible approach for modeling, and testing the significance, of the time-dependent (TD) and non-linear (NL) effects of continuous TVCs in time-to-event analyses. To account for sparsely measured TVCs, we add modeling of the effect of time elapsed since last observation (TEL), which acts as an effect modifier for the TD and NL effects. The second manuscript considers a more difficult case where the temporal association between the current hazard and the TVC measurements may depend also on the past TVC values. In this scenario, a priori choice of a simple exposure metric linking the hazard may lead to wrong conclusions. Sparse measurements of TVC values add more complexity to both the estimation of the TVC effects and the inference about them in this setting. To address these challenges, we extend the weighted cumulative exposure (WCE) framework proposed by Sylvestre & Abrahamowicz (2009) to the modeling of a sparsely measured continuous TVC in time-to-event analyses. When the TVC is sparsely measured, we propose a two-stage strategy to impute the missing TVC values, which yields less biased estimates of the weight function compared to the naive last observation carried forward (LOCF) approach. The third manuscript extends the flexible models developed in the first two manuscripts, incorporating the weighted cumulative effect of the past values of a sparsely measured continuous TVC, as well as its TD and/or NL effect(s), to allow non-parametric modeling of the baseline hazard and estimation of the resulting individual survival curves, conditional on such complex covariate effects. Using both simulated and real-life data, we also illustrated how taking into account the true complex TVC effects modified individual survival curves and enhanced their accuracy. In conclusion, the three thesis manuscripts enrich the existing toolbox for survival analysis by extending flexible modeling of the complex predictor effects to a challenging setting when the predictor of interest is a sparsely measured continuous time-varying covariate.Étant donné que des mesures répétées pour de nombreux facteurs pronostiques deviennent plus faciles à obtenir grâce, un nombre croissant d'études cliniques et épidémiologiques modernes commencent à les incorporer dans les analyses. Dans le contexte de l'analyse de survie, les facteurs de risque/pronostique mesurés de façon répétée sont habituellement incorporés en tant que covariables variant dans le temps (CVT) dans le modèle à risques proportionnels de Cox. Cependant, une modélisation adéquate des CVT continues pose souvent des défis en raison de l'incertitude sur (i) la nature de la dépendance entre le risque instantané actuel et les valeurs actuelles et/ou passées d'une CVT, (ii) la forme de la fonction de l'effet d'une CVT continue sur le log du risque instantané, et (iii) des changements potentiels de la force de l'effet en fonction du temps. De plus, dans la pratique, les mesures disponibles d'une CVT sont généralement peu fréquentes, ce qui pose des défis analytiques supplémentaires. Par conséquent, l'objectif global de cette thèse est d'améliorer la modélisation des covariables variant dans le temps en analyse de survie en abordant chacun des défis mentionnés plus haut. Les contributions originales de la thèse consistent en trois manuscrits, chacun proposant de nouvelles méthodes, les évalue par des simulations et illustre leurs applications sur des données réelles. Le premier manuscrit se concentre sur une situation relativement simple, soit lorsque le risque instantané actuel est supposé être associé uniquement à la valeur actuelle de la CVT. Le modèle à risqué proportionnels conventionnel suppose que les effets des covariables sur le risque instantané sont constants dans le temps, et que les effets des covariables continues sur le logarithme du risque instantané sont linéaires, ce qui peut souvent ne pas être le cas en pratique. Pour éviter ces hypothèses restrictives, nous proposons une approche flexible pour modéliser, et tester la signification, des effets dépendants du temps (DT) et non-linéaires (NL) des CVT continues dans les analyses de survie, et l'évaluons avec des simulations. Pour tenir compte des CVT mesurées peu fréquemment, nous ajoutons une modélisation de l'effet du temps écoulé depuis la dernière observation (TED), qui agit comme un facteur modifiant pour les effets DT et NL. Le second manuscrit considère un cas plus difficile. Dans ce scénario, le choix a priori d'une métrique d'exposition simple reliant le risque instantané peut conduire à des conclusions erronées. Des mesures peu fréquentes de la valeur d'une CVT ajoutent de la complexité à la fois à l'estimation des effets de la CVT et à l'inférence à leur sujet dans ce contexte. Pour relever ces défis, nous étendons le cadre de l'exposition cumulative pondérée (ECP) à la modélisation d'une CVT continue mesurée peu fréquemment en analyse de survie. Lorsque la CVT est mesurée peu fréquemment, nous proposons une stratégie en deux étapes pour imputer les valeurs manquantes de la CVT, qui produit des estimations moins biaisées de la fonction de pondération que l'approche naïve d'imputation reportant la dernière observation. Le troisième manuscrit étend les modèles flexibles développés dans les deux premiers manuscrits, pour permettre la modélisation non-paramétrique du risque de base et l'estimation des courbes individuelles de survie résultantes. En utilisant à la fois les données simulées, nous avons également illustré comment la prise en compte des effets complexes d'une CVT modifie les courbes individuelles de survie et améliore leur précision. En conclusion, les trois manuscrits de la thèse enrichissent la boîte à outils existants pour l'analyse de survie en étendant la modélisation flexible des effets complexes de variables prédictives au contexte difficile d'une covariable d'intérêt continue variant dans le temps mesurée peu fréquemment
- …
