In the past few years, biomarkers such as progression free survival (PFS) and time to progression (TTP), have been increasingly used as surrogate endpoints for overall survival (OS) in clinical trials in oncology. An issue occurs when clinical trials which demonstrated statistically significant treatment effect for the surrogate marker, shows no significant effect on the true outcome of interest, OS. It is possible that this lack of concordant results was due to informative censoring. Although it is known that informative censoring may bias the observed results, it is not clear what impact informative censoring has on the surrogacy of one marker in relation to a true outcome. In this thesis, we investigated how informative censoring could affect the results of a surrogate endpoint, and how would that affect the surrogacy of that endpoint. A simulation study was conducted to evaluate the impact of informative censoring on the treatment effect on TTP and the outcomes of the surrogate validation methods relative effect (RE), surrogate threshold effect (STE), and the difference between the treatment effect on TTP and on OS (IRE). The results of the simulation showed that having informative censoring for TTP will indeed bias the treatment effect on TTP as well as the results for the validation methods, RE, STE, and IRE. Hence, we conclude that the effect of informative censoring can greatly influence the ability to validate a surrogate marker, and additionally can bias the ability to determine the efficacy of a new therapy from a clinical trial using a surrogate marker as the primary outcome. / Thesis / Master of Science (MSc)

In recent years, adaptive designs in clinical trials have been attractive due to their efficiency and flexibility. Response adaptive randomization procedures in phase II or III clinical trials are proposed to appeal ethical concerns by skewing the probability of patient assignments based on the responses obtained thus far, so that more patients will be assigned to a superior treatment group. General response-adaptive randomizations usually assume that the primary endpoint can be obtained quickly after the treatment. However, in real clinical trials, the primary outcome is delayed, making it unusable for adaptation. Therefore, we utilize surrogate and primary endpoints simultaneously to adaptively assign subjects between treatment groups for clinical trials with continuous responses. We explore two types of primary endpoints commonly used in clinical tirials: normally distributed outcome and time-to-event outcome. We establish a connection between the surrogate and primary endpoints through a Bayesian model, and then update the allocation ratio based on the accumulated data. Through simulation studies, we find that our proposed response adaptive randomization is more effective in assigning patients to better treatments as compared with equal allocation randomization and standard response adaptive randomization which is solely based on the primary endpoint.

Most statistical methods to design and analyze randomized controlled trials with time-to-event data, and synthesize their results in meta-analyses, use the hazard ratio (HR) as the measure of treatment effect. However, the HR relies on the proportional hazard assumption which is often violated, especially in cancer trials. In addition, the HR might be challenging to interpret and is frequently misinterpreted as a risk ratio (RR). In meta-analysis, conventional methods ignore that HRs are estimated over different time supports when the component trials have different follow-up durations. These issues also pertain to advanced statistical methods, such as network meta-analysis and surrogate endpoints validation. Novel methods that rely on the difference in restricted mean survival times (RMST) would help addressing these issues. In this dissertation, I first developed a Bayesian network meta-analysis model using the difference in RMST. This model synthesizes all the available evidence from multiple time points and treatment comparisons simultaneously through within-study covariance and between-study covariance for the differences in RMST. I proposed an estimator of the within-study covariance and estimated the model under the Bayesian framework. The simulation studies showed adequate performance in terms of mean bias and mean squared error. I illustrated the model on a network of randomized trials of second-line treatments of advanced non-small-cell lung cancer. Second, I introduced a novel two-stage meta-analytical model to evaluate trial-level surrogacy. I measured trial-level surrogacy by the coefficient of determination at multiple time points based on the differences in RMST. The model borrows strength across data available at multiple time points and enables assessing how the strength of surrogacy changes over time. Simulation studies showed that the estimates of coefficients of determination are unbiased and have high precision in almost all of the scenarios we examined. I demonstrated my model in two individual patient data meta-analyses in gastric cancer. Both methods, for network meta-analysis and surrogacy evaluation, have the advantage of not involving extrapolation beyond the observed time support in component trials and of not relying on the proportional hazard assumption. Finally, motivated by the common misinterpretation of the HR as a RR, I investigated the theoretical relationship between the HR and the RR and compared empirically the treatment effects measured by the HR and the RR in a large sample of oncology RCTs. When there is evidence of superiority for experimental group, misinterpreting the HR as the RR leads to overestimating the benefits by about 20%. / 2026-02-08T00:00:00Z