Methods of accelerated life testing (ALT) are widely used in reliability theory for estimating lifetime of technical devices. To this task, items are exposed systematically to higher stress levels of, e.g. temperature, voltage or pressure. This approach is more efficient by providing more failures in shorter observation time and information for higher stress levels is used to estimate lifetime under normal conditions.

We propose to use these ideas in the epidemiology of chronic diseases by conceptualizing the diagnosis of a chronic disease as exposing a person to a higher stress level, thus potentially shortening its residual lifetime, or, equivalently, accelerating its time to death. We use the tampered random variable model of Degroot and Goel (1979) and Gompertz distributions to model mortality from type 2 diabetes using data from the population-based CARLA cohort. The TRV model correctly accounts for the semi-competing risk structure in the data, allows entry into the cohort at higher ages, and uses information from prevalent as well as incident cases. In addition, using parametric distributions offers reporting model results on the original time, rather than on the hazard scale. In an extension of the model we also allow the age at diabetes diagnosis to be observed not exactly, but only in an interval. Model parameters can be estimated straightforwardly by maximum likelihood, and we give some preliminary results of a simulation study showing our approach working well.

The Hazard Ratio (HR) is often reported as the main causal effect when studying survival data. Despite its popularity, the HR suffers from an unclear causal interpretation due to a built-in selection bias. While alternative approaches exist, the HR remains the most popular measure used by practitioners, and therefore, analysis approaches directly targeting a causally interpretable HR are of interest. A recently proposed alternative is the causal HR, defined as the ratio between hazards across treatment groups among the study participants that would have survived regardless of the assigned study group. We discuss the challenge in identifying the causal HR from the observed data and present a sensitivity analysis approach for identification in randomized controlled trials utilizing a working frailty model. We further extend our framework to adjust for potential confounders using inverse probability of treatment weighting. We present a Cox-based and non-parametric kernel-based estimation under right censoring. We study the finite-sample properties of the proposed estimation methods through simulations and illustrate the utility of our framework using two real-data examples.

Propensity score matching has become a popular method for estimating causal treatment effects in nonrandomized studies. However, for time-to-event outcomes, the estimation of hazard ratios based on propensity scores can be challenging if omitted or unobserved covariates are disregarded. Not accounting for such covariates could lead to heavily biased treatment estimates. Researchers often do not know whether (and, if so, which) covariates will induce this bias. To address this issue, we extended a previously described method, “Dynamic Landmarking”, which was originally developed for randomized trials. By simulation we show, that “Dynamic Landmarking” provides a good visual tool for detecting biased treatment estimates also in propensity score matched data. The underlying approach will be applied to a real world data set from cardiac surgery.

Background

The analysis of multivariate serological data, i.e., Type I interval-censored or current status data coming from blood serum samples tested for the presence of antibodies against multiple pathogens, gained attention in recent years. More specifically, refinements towards accounting for individual heterogeneity in the acquisition of infections, non-immunizing infection dynamics and more complicated association structures have been proposed and studied in detail. Despite the common use of a so-called threshold approach to classify individuals as seronegative or -positive depending on a continuous antibody titer measurement for each pathogen under study, the subjective choice of a single or even multiple thresholds is a strong limitation. In this work, we focus on directly using individual-level bivariate continuous antibody titer data to estimate parameters related to the occurrence of two pathogens.

Methods

We consider a Bayesian bivariate mixture approach to model continuous antibody titer data on two pathogens in the presence of individual heterogeneity and to study the implied association in the acquisition of these two pathogens. Our approach extends a common frailty approach for paired current status data, which is subject to potential misspecification of thresholds as previously mentioned.

Data application

We fitted the aforementioned model to bivariate serological data on varicella-zoster virus (VZV) and parvovirus B19 (PVB19). Given recent evidence of possible reinfections with PVB19, we investigated processes of waning of humoral immunity by allowing for an age-dependent change in mean antibody titer concentration. We discuss these humoral immunity processes in more detail and consider different model choices in view of such processes.

Results and conclusions

The estimated seroprevalence for PVB19 is characterized by a steep increase with increasing age, following infections among young children, followed by a decrease between the age of 20 to 40 years after which the seroprevalence increases again. Moreover, the evolution of the mean antibody titer concentrations is rather constant across age groups, indicating that despite a decay in humoral immunity at the individual-level, population-level mean antibody titer values remain unchanged because of reinfections with PVB19 among 20-40 years old. Given the risk of spontaneous abortion after PVB19 infection during pregnancy, waning of humoral immunity in 20-40 years could be responsible for an excess of miscarriage and fetal death. For VZV, the seroprevalence is monotonically increasing, indicating that varicella infection is responsible for high levels of humoral immunity persisting for life. The mean antibody levels show a slight decrease with increasing age among seropositive individuals, however, not to an extent that seroprotection is not ensured for life. In general, based on our analyses, we showed that the mixture model provides additional insights concerning waning of IgG antibodies as compared to more traditional frailty approaches while the model is sufficiently flexible to capture observed dynamics in IgG antibodies. Furthermore, the model accounts for association in the acquisition of the pathogens under study through the specification of random effects termed frailties to explicitly link our approach to survival models that have been used in the past.