Institute of Biostatistics and Registry Research
Biostatistics applies mathematical theory and statistical methods to the collection, analysis and interpretation of biological data, in particular relating to human biology, health, and medicine. Biostatistics is a key component of biomedical research and is essential for the interpretation of results from clinical studies.
The Institute of Biostatistics and Registry Research develops and applies methodological and statistical approaches to biological data to study the causes of disease as well as the prevention and treatment of diseases. In addition, the institute is setting up a registry research program. This program involves all aspects of using registry data for research purposes, including the design of disease registries as well as clinical studies using registry data, expanding methodology to conduct registry linkages for health and health care evaluations, and conducting collaborative research in this field.
Members of the institute have particular expertise in the design and analysis of observational clinical and population-based studies, including the modelling and prediction of risks for late effects after cancer treatment, the evaluation of predictive markers for personalized medicine, the assessment of long-term health problems after exposure to ionizing radiation for diagnostic and therapeutic purposes,
record linkage approaches to combine data from several registries, study designs using registry data, and the summary and translation of evidence from epidemiological studies into recommendations for individual risk assessment or clinical practice.
The institute closely collaborates with clinicians and other researchers of the Brandenburg Medical School and beyond on various clinical studies. Members of the institute provide statistical and methodological expertise on diverse topics from all areas of biomedical research, including the design, analysis and reporting of biomedical studies.
Moreover, the institute offers courses in medical statistics and related subjects for students of the Brandenburg Medical School. The courses provide an introduction into design aspects, methods of summarizing and presenting data, risk estimation, calculation of confidence intervals, hypothesis testing, and correlation for the assessment of association.
We are looking for a PhD student in Biostatistics with good German language skills for a project on predictive markers for personalized breast cancer treatment. Details can be found here.
Van der Willik KD, Hauptmann M, Jóźwiak K, Vinke EJ, Ruiter R, Stricker BH, Compter A, Ikram A, Schagen SB. Trajectories of cognitive performance prior to cancer diagnosis: a population-based study. In press.
Teepen JC, Kok JL, Kremer LC, Tissing WJ, Van den Heuvel-Eibrink MM, Loonen JJ, Bresters D, Van der Pal HJ, Versluys B, Van Dulmen-den Broeder E, Nijsten T, Hauptmann M, Hollema N, Dolsma WV, Van Leeuwen FE, Ronckers CM, DCOG-LATER Study Group. Long-term risk of skin cancer among childhood cancer survivors: a DCOG-LATER cohort study. J Natl Cancer Inst 2019; 111(8): 845–853.
Statistical methods for the evaluation of late health effects from therapeutic radiation exposure
Since the number of cancer survivors is rapidly increasing, it is important to understand the late effects of cancer treatment, particularly second primary cancers caused by exposure of healthy tissue to radiation. Accurate risk assessment is essential for predictions to be included in decision making at diagnosis and after treatment. Incorporation of dose distributions is currently not standard in epidemiologic studies of radiotherapy-related second cancer risk. It is expected to yield more efficient and less biased estimates of the dose-response relationship as well as better risk predictions for clinical use. Within a project funded by the Dutch Cancer Society, PhD student Sander Roberti and scientific programmer Dr. Viet Nguyen characterize statistical methods incorporating dose distributions in the organ at risk for a second tumor with regard to efficiency and bias, and describe risk predictions for second cancers following current radiotherapy, calculated by these statistical methods. They will use data from the BRIGHT study of breast cancer among Hodgkin lymphoma survivors for which the three-dimensional distribution of radiation doses has been retrospectively estimated by Russell et al. (2017). Three-dimensional dose distribution data will also be used within a DCOG LATER study on meningioma among childhood cancer survivors by Kok et al (2019). Collaborators are Dr. Nicola Russell and professor Flora van Leeuwen from the Netherlands Cancer Institute in Amsterdam (The Netherlands), Dr. Cécile Ronckers and Professor Leontien Kremer from the DCOG LATER consortium of the Princess Máxima Center for Pediatric Oncology in Utrecht (The Netherlands), Dr. Ruth Pfeiffer from the National Cancer Institute in Bethesda (USA) and other international partners.
Design and statistical analysis of clinical studies for the evaluation of markers for personalized medicine
To improve survival after a cancer diagnosis, biomarkers are urgently needed to identify patients who do and do not derive benefit from certain treatments, in order to personalize treatment to the individual patient. Few such biomarkers are currently used in clinical practice, which could be partly due to the fact that promising candidate markers do not survive the rigorous first steps of clinical evaluation. These early studies are often small and not randomized. Subsequent large randomized clinical trials are then not conducted to confirm a biomarker and introduce it into clinical practice. In a project funded by the Dutch Cancer Society, a PhD student and scientific programmer Dr. Viet Nguyen assess statistical designs and methods for the evaluation of predictive markers in observational clinical studies or trials with archived specimens. The designs include case-only and hybrid approaches, and we employ additive and multiplicative models. We investigate the required sample size and statistical power as well as other operational characteristics based on simulated data and application to data on markers of DNA repair deficiency and chemotherapeutic treatment of breast cancer. Several studies including those by Vollebergh et al (2011), Schouten et al (2015) and Puppe et al (2019) suggest that a marker for a deficiency to repair DNA double-strand breaks (“BRCA-like”) can identify women who derive substantial benefit from high-dose chemotherapy. Other studies are underway to explore whether the marker also identifies subgroups of breast cancer patients particularly sensitive to other chemotherapeutic treatments, including The German ADAPT-TN trial and the Finnish FinXX trial. For this project, we collaborate with professor Sabine Linn from the Netherlands Cancer Institute in Amsterdam (The Netherlands), professor Rita Schmutzler from the University of Cologne (Germany), professor Heikki Joensuu from the University of Helsinki (Finnland) and other international partners.
Cancer risk after medical ionizing radiation exposure
Diagnostic and therapeutic sources are the major contributors to ionizing radiation exposure of the population, and these exposures are predicted to increase substantially in the future. Diagnostic computed tomography (CT) scans have increased in most Western countries during the last 20-30 years, including the Netherlands (Meulepas et al 2017). At the same time, CT scans deliver a relatively high dose of radiation compared with other diagnostic modalities. However, cancer risks from such doses of radiation are currently not well understood. Funded by the European Commission and Worldwide Cancer Research, we evaluated subsequent risk of cancer due to the radiation exposure in a large retrospective cohort study of Dutch children who underwent a CT scan (Meulepas et al 2014), and found a dose-response relationship for brain tumors (Meulepas et al 2019). For the “Epidemiological study to quantify risks for pediatric computerized tomography and to optimize doses” (EPI-CT), funded by the European Commission, our data are being pooled with similar data from 8 other European countries and are jointly analyzed. Details on the design of the pooled cohort including about one million children have been published (Bernier et al 2019).
We participate in the study „Implications of medical low dose radiation exposure“ (MEDIRAD), funded by the European Commission. Within this large consortium conducting various studies on the health effects of medical radiation exposure, we update the pooled cohort data collected within the EPI-CT project (see above) in order to increase the number of cases and the associated statistical power.
Indirect adjustment for unmeasured confounders in different study designs
Discussions about the strength of evidence derived from randomized versus observational studies have started a long time ago (Vandenbroucke 2004, Vandenbroucke 2011). This discussion is ongoing until today. For example, reports by the Agency for Healthcare Research and Quality on the assessment and correction of bias in observational studies and by the European Medicines Agency (EMA) on the use of patient registries for regulatory purposes have been controversially discussed, among others by the German Institute for Quality and Efficiency in Health Care (IQWiG). On the other side, Cochrane has compared estimates from observational and randomized studies and found little difference. As a consequence of these discussions and investigations, quantitiative bias assessment has gained importance in observational studies. The application of corresponding methods has been facilitated by the textbook of Lash et al 2009 and publications by Ding and VanderWeele 2016 and VanderWeele and Ding 2017, among others.
In studies of medical interventions as the determinants of disease, indication bias is a particularly challenging problem of observational studies. Since details of the indication are often not recorded and randomized studies are often impossible, methods to assess and control for unmeasured confounding by indication are essential.
We have developed and applied methods to externally assess and correct for bias from unmeasured confounders, including indication, in cohort and case-control studies of medical radiation exposure and cancer risk (Lubin et al 2018, Brenner et al 2017, and Meulepas et al 2016). We are currently extending these methods to other study designs.
International Commission on Radiological Protection (ICRP)
As a member of the committee on biological effects of the ICRP, Michael Hauptmann is currently involved in a study on factors governing the individual response of humans to ionizing radiation. There are many factors that, to different degrees, influence the responses of individual people to radiation (Rajaraman et al 2018). In addition to the obvious factors of radiation quality, dose, dose rate and the tissue (sub)volume irradiated, determining factors include, among others, age and sex, life style (e.g. smoking, diet, and possibly body mass index), environmental factors, genetics and epigenetics, stochastic distribution of cellular events and systemic comorbidities such as diabetes or viral infections. Genetic factors are commonly thought to be a substantial contributor to individual response to radiation. This study reviews the current evidence from studies of late adverse tissue reactions after radiotherapy in potentially sensitive groups, including data from functional assays, candidate gene approaches, and genome-wide association studies.