|June 2nd||11:00-12:00 at Zoom||| Christian Bressen Pipper | Senior Statistical Advisor, Biostatistics and Pharmacoepidemiology, LEO Pharma A/S |
A statistical journey towards good answers
|ABSTRACT: I think we can all agree that the biggest merit of statistics is its ability to robustly summarize empirical evidence to support decision making. When I started my career as a statistician roughly 20 years ago this was done focusing on flexible statistical models and efficient estimation in order to provide correct answers and precise quantifications of treatment (or other) effects no matter how oddly data behaved. In those days not too much thought was given as to whether the resulting answers we gave made sense to our non-stat collaborators. Luckily for everyone this has changed towards focusing on defining and targeting relevant treatment effects utilizing tools like causal reasoning and the estimand framework. In this talk I will exemplify how this has changed my research focus all the way from semiparametric inference for multivariate time to event data to methods for providing relevant quantifications of treatment interventions in randomized controlled trials.|
|May 19th||11:00-12:00 at Zoom||| Henrik Støvring | PhD, DMSc, Department of Public Health, Aarhus University |
Estimation of prescription duration from dispensing dates - challenges and new opportunities with the Waiting Time Distribution
|ABSTRACT: Determining the duration of treatment associated with a redeemed prescription as observed in electronic pharmacoepidemiologic databases is a fundamental problem in pharmacoepidemiologic research. Traditionally researchers have relied on simple decision rules regarding occurrence of medication dispensing in a fixed time window, although such decision rules are prone to bias and lack theoretical underpinnings. In this presentation I will review some of the fundamental challenges to estimation of prescription durations and introduce the parametric Waiting Time Distribution (WTD), which allows estimating prescription durations using regression-like statistical techniques without relying on simplistic decision rules. Finally, I will present some of our recent developments of the method to improve its statistical efficiency and integration with models for exposure and outcome in pharmacoepidemiologic case-control studies.|
|14:00-15:00 at Zoom||| Dr. Oskar Hansson and Dr. Niklas Mattson-Carlgren |
Recent discoveries in biomarkers for Alzheimer's disease
|Feb 2nd||11:15-11:45at Zoom||| Xiaomeng Zhang |
| PhD Student, Centre for Global Health Research, Usher Institute, University of Edinburgh, UK |
Genetically determined colorectal cancer risk is associated with multiple health outcomes: Phenome-wide Association study in the UK Biobank
|ABSTRACT: Associations between colorectal cancer (CRC) and other health outcomes have been widely investigated. However, the causal relationships between various health outcomes and CRC risk are poorly understood.
Xiaomeng Zhang will present her recent work, where she explored association between multiple outcomes and CRC under the phenome-wide association framework including phenome-wide association study (PheWAS) and tree-structured phenotypic model (TreeWAS). Analysis was done in 339,198 unrelated White British individuals from the UK Biobank cohort. A polygenic risk score (PRS) generated from two CRC genome-wide association studies was taken as a proxy of CRC risk. A range of sensitivity analyses was performed to test the robustness of the results and the Danish Disease Trajectory Browser (DTB) was searched to replicate the associations under the observational settings.
Eight out of 1356 phenotypes and 17 out of 8773 nodes were identified to be associated with genetically predicted CRC by PheWAS at false discovery rate q<0.05 and TreeWAS at a posterior probability greater than 0.95, respectively. All detected associations were from neoplasms and digestive system disease groups (e.g. benign neoplasm of colon and diverticular disease). Sensitivity analyses supported an inference that associations between other anaemia, type 2 diabetes, renal failure and bacterial infection and predisposition to CRC were driven by CRC cases.
In summary, these findings suggest that benign or pre-malignant colorectal neoplasms, neoplasms of unspecified sites, and diverticular disease shared the biological pathway with CRC. Other anaemia, type 2 diabetes, renal failure and bacterial infection implicated co-occurrence with CRC. Further, Xiaomeng plans to expand the current analysis to 136 blood and urine biomarkers available for the UK Biobank cohort.
|Apr 20th||13:00-14:00 in Zoom||| Maria Timofeeva |
| Assistant Professor, EBB, SDU |
Gene Expression of Cancer Risk and Survival
|ABSTRACT: Over the last several decades thousands of genetic variants have been identified to be associated with common diseases and traits including cancers. These findings provide a unique way to discover the causal biological mechanisms without no prior biological hypothesis and already have given us important insights into underlying bases of many cancer types. Many of these genetic variants are also expression quantitative traits loci, which means they correlate with gene expression. Gene expression is the process by which the information in a gene is converted into a functional gene product (RNA and proteins). It has been suggested that many of the genetic variants influence disease risk though modulating gene expression. Lifestyle choices and environment have an impact on gene expression too, bringing gene regulation and expression to the forefront of interactions between nature and nurture. Understanding causal relationships between gene expression and cancer outcomes has potential for identification of novel drug targets as well as developing targeted screening programmes for the high-risk populations.
However, analysis of gene expression has not been always possible in populational based studies. High cost of whole genome analysis, poor availability of biological samples collected prior to disease onset and complex nature of gene expression make it extremely difficult. This is where twin design offers a powerful tool to study association between gene expression and cancer outcomes.
In my talk I will demonstrate how bringing together genetic and gene expression association studies can shed light on function of candidate genes using an example of a recent genome-wide association study on colorectal cancer. I will also present my vision how some of these questions can be addressed using the Danish twin registry.
|Jan 15th||11:00-12:00 in room 4.39||| Dennis Lund Hansen & Henrik Frederiksen |
| Department of Hematology, OUH |
Wampires and hematology
|ABSTRACT: Hematology concerns blood and lymphatic diseases, sometimes including vampires. We have recently proposed that leukemia could be an explanation for the vampire victims symptoms in the novels Carmilla and Dracula. We will tell about this and other ongoing epidemiological projects in our department.|
|Dec 3rd||9:30-10:15 in room 4.39||| Simon Bang Kristensen |
| PhD Fellow, Section of Biostatistics, Aarhus University |
Bivariate logistic mixed effect regression models based on latent variables
|ABSTRACT: Bivariate observations of binary and ordinal data arise frequently in clinical trials and the cognitive sciences. Consider for example an efficacy-toxicity study for some drug, in which efficacy is binary and toxicity is measured as side-effects on an ordinal scale. While robust methods offer the possibility to evaluate drug effects taking into account the correlation between efficacy and toxicity, a bivariate modelling approach is required in scenarios where one is interested in aspects of the marginal distributions in themselves along with the association between the two. For example, we may be interested in efficacy and toxicity as separate outcomes but also wish to estimate the probability of benefiting from a positive treatment outcome while not experiencing a high degree of side-effects as a function of dosage.
We consider methods for constructing such bivariate models with logistic marginals and propose a model based on the Ali-Mikhail-Haq bivariate logistic distribution. We motivate the model as an extension of that based on the Gumbel type 2 distribution as considered by other authors and as a bivariate extension of the logistic distribution which preserves certain natural characteristics.
Basic properties of the obtained model are studied and the proposed method is exemplified by analysis of a cognitive experiment of visual recognition and awareness.
|Oct 11th||11:00-12:00 in room 4.39||| Dr Mauro Laudicella |
| DaCHE, SDU |
20 Years of Falling Hospital Rates: What Impacts Can We Expect?
|ABSTRACT: In the past twenty years, hospital mortality rates have fallen dramatically in many high-income countries due to constant improvements in hospital quality of care and investments in new technologies. I will present two key studies investigating the impact of the decline in hospital mortality rates on the demand for unplanned acute care in England; I will then discuss their implications for the Danish Health System and the scope for further investigations. In the first study, I examined the impact of reducing hospital mortality rates on hospital readmissions occurring after 30 days from first admission.
The study found evidence of a negative relationship between hospital performance in mortality and readmissions: hospital that were more successful in reducing their mortality rates experienced greater readmissions rates due to a larger share of frail patients surviving the first admission and thus at risk of being readmitted as compared to other hospitals. In the second study, I examined the impact of reducing hospital mortality rates nation wide and over a ten-years time period on the number of emergency admissions experienced by patients one year after the first admission.
The study found evidence that one third of the growth in the demand for emergency admissions can be explained by the increasing number of patients surviving their first admission overtime, but increasingly frail and at risk of being readmitted to the hospital within one year.
1. Laudicella, M., P. Li Donni, and P.C. Smith, Hospital readmission rates: signal of failure or success? J Health Econ, 2013. 32(5): p. 909-21.
2. Laudicella, M., et al., Do Reduced Hospital Mortality Rates Lead to Increased Utilization of Inpatient Emergency Care? A Population-Based Cohort Study. Health Serv Res, 2018. 53(4): p. 2324-2345.
|June 6th||10:00-12:00 in room 4.39||
Research on tattoos and risk of cancer| Dr Milena Foerster |
| The International Agency for Research on Cancer, Lyon, France |
| Dr Jørgen Serup |
| Bispebjerg Hospital, Copenhagen University, the Tattoo Clinic |
| Karina Friis |
| DEFACTUM Research, Aarhus University |
| Signe B. Clemmensen | | EBB, SDU |
|ABSTRACT: Studies regarding the health risks of tattoos are limited. Epidemiological studies of tattoos and association with cancer risk are basically non-existent as exposure data are hard to get. Several researchers are concerned about the lack of knowledge because the trend of tattooing is rapidly increasing: 15-20% of adult Danes are estimated to be tattooed.
This seminar has been arranged as part of the first meeting in a discussion group comprised of researchers who are working on or are initiating studies about the exposure to tattoo ink and possible association with risk of cancer. The 4 groups (see above) will present themselves and their upcoming studies and there will be time for debate between each presentation.
|May 2nd||10:00-11:00 in room 4.48||| Dr Noreen Goldman |
| Professor at Princeton University |
Predicting Survival of Older Adults
|April 9-10||9 Apr 13:00 - 10 Apr 13:00 in room 15-0.55||
2-Day Meeting in the Danish Society of Theoretical Statistics - DSTSSpeaker, titles, and abstracts for the presentations
|Jan 15th||9.00-13.00||| Claire Steves |
| Professor at Department of Twin Research and Genetic Epidemiology |
| King's College London |
Multi-morbidity and frailty: Insights from Twins UK