Novel methods for complex time-to-event and longitudinal data in medical studies

In complex diseases individuals are often at risk of several types of possibly semi-competing events and may experience recurrent symptomatic episodes. This complex disease course makes it challenging to define target estimands for clinical trials. While composite endpoints are routinely adopted, recent innovations involving the win ratio and other methods based on ranking the disease course have received considerable attention. We emphasize the usefulness of multistate models for addressing challenges arising in complex diseases, along with the simplicity and interpretability that comes from defining utilities to synthesize evidence of treatment effects on different aspects of the disease process. Robust variance estimation based on the infinitesimal jackknife means that such methods can be used as the basis of primary analyses of clinical trials. We illustrate the use of utilities for the assessment of bleeding outcomes in a trial of cancer patients with thrombocytopenia. This is joint work with Alexandra Buhler, Richard Cook and Jerry Lawless. 

Case-control study designs are commonly used in biomedical studies because they require less time and expense than a prospective cohort study to conduct satisfactory analyses. Such designs are applicable for finding risk factors of a disease when study subjects can be classified as cases vs. controls according to the status of being diseased or disease-free. When adopting such a sampling design, survivor bias could form a serious problem which may lead to biased analysis results. Existing approaches typically treated cases and controls in binary form without specifying age at incidence of disease, and assumed the prevalent cases are sampled with length bias in stationary models. Instead of binary disease outcome, we consider age-specific risk outcome and propose a composite likelihood approach which handles survival bias in either stationary or non-stationary models. A data analysis is presented using data from the Alzheimer Biomarkers Consortium - Down Syndrome (ABC-DS) Study to illustrate the applicability of the proposed methods. 

Joint modelling is an attractive approach for prediction of longitudinally observed outcomes and causal inference, based on time-varying markers. Unfortunately, the application and development of this methodology is still hampered by speed and optimization problems in model estimation and above all in reaching sufficient scale for realistic practical application, especially through restrictions on the number of longitudinally recorded parameters which can be used for prediction of outcomes. Integrated nested Laplace approximations can address some of these problems by implementing a deterministic form of estimation which is both speedy and accurate, while still facilitating prediction application. We discuss application in a study on Duchenne muscular dystrophy at the Leiden University Medical Centre where multiple longitudinally tracked proteins must be combined to predict a bounded-range time-varying functional outcome score. A joint model is formulated which utilizes a beta-density outcome model for the time-dependent outcome, linked to the multiple protein predictors using the so-called scaled-predictor parametrization. Estimation through the integrated nested Laplace approach and results from a data analytic application are demonstrated.