Development and application of statistical models to evaluate potential treatment effects in observational COVID-19 studies
Final Report Abstract
We present the progress and findings of our study on the methodological evaluation of observational studies for COVID-19 treatment effectiveness. Given the complexity of clinical endpoints, including varying oxygen support states, intensive care unit (ICU) admissions, discharge, and mortality, observational studies require careful methodological considerations to ensure robust and unbiased results. Unlike randomized controlled trials (RCTs), where treatment assignment is random, observational studies are prone to biases such as timedependent confounding, immortal time bias, and competing risks, which can distort treatment effect estimates and lead to misleading conclusions. To address these challenges, we developed and applied advanced statistical models to improve evidence-based decision-making in non-randomized studies. Our research focused on systematically identifying and mitigating methodological biases through a comprehensive literature review, evaluating statistical approaches, and implementing novel frameworks to enhance the validity of observational findings. We critically examined the limitations of existing studies, highlighting common pitfalls in analyzing real-world data and their implications for clinical decision-making. One of the key aspects of our work was the application of target trial emulation, a framework that mimics randomized trials using observational data. This approach allows researchers to structure data analysis as if a hypothetical clinical trial were conducted, thereby reducing bias and improving causal inference. Additionally, we implemented multi-state modeling, a technique designed to account for complex patient trajectories across different health states, providing a more nuanced understanding of disease progression and treatment effectiveness. Our findings demonstrate that a combination of target trial emulation and multi-state modeling significantly enhances the reliability of observational studies. By applying these methodologies to real-world COVID-19 data, we were able to improve the accuracy of treatment effect estimates, providing valuable insights into the efficacy of different therapeutic interventions. This research not only contributes to the field of epidemiology and biostatistics but also has practical implications for policymakers, healthcare providers, and researchers who rely on observational data to inform treatment guidelines and public health strategies. Ultimately, this study underscores the importance of rigorous statistical modeling in nonrandomized studies, emphasizing the need for transparent methodologies and well-defined analytical frameworks. The approaches developed in this project can be applied beyond COVID-19 research, extending to various domains of clinical epidemiology where observational data play a crucial role in assessing treatment effectiveness.
Publications
-
How to Quantify and Interpret Treatment Effects in Comparative Clinical Studies of COVID-19. Annals of Internal Medicine, 174(5), 731.
Hazard, Derek; Kaier, Klaus; von Cube, Maja; Lambert, Jerome & Wolkewitz, Martin
-
Methodological evaluation of bias in observational coronavirus disease 2019 studies on drug effectiveness. Clinical Microbiology and Infection, 27(7), 949-957.
Martinuka, Oksana; von Cube, Maja & Wolkewitz, Martin
-
Response to “Overlooked Shortcomings of Observational Studies of Interventions in Coronavirus Disease 2019: An Illustrated Review for the Clinician” by Tleyjeh et al. Open Forum Infectious Diseases, 9(1).
Wolkewitz, Martin & Martinuka, Oksana
-
Target Trial Emulation Using Hospital-Based Observational Data: Demonstration and Application in COVID-19. Life, 13(3), 777.
Martinuka, Oksana; Cube, Maja von; Hazard, Derek; Marateb, Hamid Reza; Mansourian, Marjan; Sami, Ramin; Hajian, Mohammad Reza; Ebrahimi, Sara & Wolkewitz, Martin
-
Target trial emulation with competing risk analysis: Demonstra tion of methodological approach to assess treatment effective ness in patients hospi talised with corona virus disease 2019. 33rd European Congress of Clinical Microbiology and Infectious Diseases (ECCMID), Copenhagen, Denmark on 15 – 18 April 2023
Martinuka O., Hazard D., … & Wolkewitz M.
-
Target trial emulation with multi-state model analysis to assess treatment effectiveness using clinical COVID-19 data. BMC Medical Research Methodology, 23(1).
Martinuka, Oksana; Hazard, Derek; Marateb, Hamid Reza; Maringe, Camille; Mansourian, Marjan; Rubio-Rivas, Manuel & Wolkewitz, Martin
-
Target trial emulation with multi-state model analysis: Demonstra tive assessment of treatment effective ness for hospitalised COVID-19 patients. Survival Analysis for Jun ior Researchers (SAfJR) Ulm, Germany, 13 – 15 September, 2023
Martinuka O., Hazard D., … & Wolkewitz M.
-
Methodological biases in observational hospital studies of COVID-19 treatment effectiveness: pitfalls and potential. Frontiers in Medicine, 11.
Martinuka, Oksana; Hazard, Derek; Marateb, Hamid Reza; Mansourian, Marjan; Mañanas, Miguel Ángel; Romero, Sergio; Rubio-Rivas, Manuel & Wolkewitz, Martin
-
Multi-state models as statistical tools for pandemic prepared ness: length of stay estimation, clinical prediction, and treat ment effectiveness. 2nd National Conference on Infectious Disease Modeling (org. MODEL LIERUNGSNETZ FÜR SCHWERE INFEK TIONSKRANKHEITEN (MONID)), German Na tional Academy of Sci ences Leopoldina, Halle (Saale), Germany, 13 – 15 March 2024
Hazard D., Mar tinuka O. & Wolkewitz M.
-
Target trial emulation with competing risk analysis using obser vational hospital data from COVID-19 patients. 45th Annual Conference of the International Socie ty for Clinical Biostatistics (ISCB45) Thessaloniki, Greece, 21 - 25 July, 2024
Martinuka O., Hazard D., … & Wolkewitz M.
