Open Collections

Li, Lingyi

University of British Columbia

Objectives: This dissertation aims to compare the risk of multiple sclerosis (MS) in anti-tumor necrosis factor alpha (TNFα) users with nonusers among patients with rheumatic disease (RD) or inflammatory bowel disease (IBD). It also aims to uncover methodological biases in existing research and explore statistical strategies to address these biases, with a particular focus on the issue of sparse data bias. Methods: Utilizing population-based health administrative data from four Canadian provinces, a nested case-control study was conducted among patients with RD and IBD (2000 to 2018). Any anti-TNFα dispensations in the two years prior to the index date (MS onset) were identified. Causal directed acyclic graphs (cDAGs) were utilized to illustrate biases like confounders, mediators, and collider-stratification bias, which may influence the relationship between anti-TNFα therapy and MS risk. Advanced statistical techniques were applied to mitigate sparse data biases. These techniques included Firth bias adjustment, data augmentation, Markov Chain Monte Carlo (MCMC), Least Absolute Shrinkage and Selection Operator (LASSO), and Ridge regression, and their results and performance were compared against traditional models via simulation studies. Results: 1) The study found that anti-TNFα therapy was associated with an increased risk of MS in RD patients (pooled incidence rate ratio [IRR]=2.05 [95% confidence interval {CI}, 1.13-3.72]) after adjusting for potential confounders. The number needed to harm was calculated at 2,268 for RD patients. While an increased risk was also observed in IBD patients, the CI was wider (pooled IRR=1.35 [95% CI, 0.70-2.59]). Sensitivity analyses and the computation of E-values were conducted to strengthen the findings. 2) When applying various statistical methods to address sparse data issues, data augmentation and MCMC approaches demonstrated superior performance in bias and mean squared error reduction in simulation studies. Conclusions: The use of anti-TNFα was associated with an increased risk of MS compared with nonusers, especially among patients with RD. The innovative use of cDAGs offers a new perspective on assessing causal relationships and addressing methodological challenges in pharmacoepidemiology. Data augmentation and MCMC approaches should be considered in pharmacoepidemiologic studies with sparse data to avoid drug effect overestimation, which can influence clinical decision-making and public health policies.

Text

2024-04-11

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/87761

Experimental Medicine

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/