Statistical methods for longitudinal and survival data

点击翻译按钮获取中文摘要

The focus of the proposed research is the development of new statistical methods for the analysis of time to event (survival) data and longitudinal/clustered outcomes. The objectives of the research program are: to compare the effects of accrual on sample size/power in cluster randomized trials that measure survival outcomes; to develop approaches for estimation of the intraclass correlation coefficient (ICC) for survival time outcomes in the presence of censoring; to develop statistical methods for interval estimation for standardized event ratios (SERs) when information on follow up time is available; to investigate regression modelling for progression-free and overall survival. In cluster randomized trials, groups of individuals (e.g. families) are randomized to treatment or intervention groups and the correlation among individuals within groups must be taken into account. New sample size/power calculation approaches in this context will be developed to include patient recruitment (accrual) information. The ICC is estimated to quantify correlation among observations within a cluster where clusters comprise individuals found in intact groups such as families, as in estimating heritability in genetic studies. When outcomes of interest are survival times, censored observations are missing exact event times. For clustered survival times, estimation of ICC is complicated by these missing survival times. New methods for estimating ICC that include both censored and uncensored observations will be developed. Standardized incidence ratios and standardized mortality ratios are examples of standardized event ratios (SERs). These ratios provide comparisons of observed versus expected events of interest. Historically, SER confidence intervals have focused on assuming a Poisson distribution for observed counts and assuming that expected counts are fixed. New methods for SER interval estimation will be developed for the scenario where follow up times are available for individuals and it is of interest to compare centres with respect to the event of interest, as in comparing standardized transplant ratios among kidney transplant centres. Often both overall survival (OS) (time to death) and progression-free survival (PFS) (time to disease progression) are of interest in clinical trials. For health economic analyses, individual clinical trial outcomes are not always available and published PFS and OS curves are used to recreate data for cost benefit analyses. Regression models that can investigate treatment effects while controlling for covariates will be developed. In addition, the impact of availability of only published PFS and OS summary curves rather than individual-level data will be investigated. The requested funding will be used primarily to support students. Recent awards recognizing mentorship speak to the quality of student supervision that will be provided.