A DAG-based comparison of interventional effect underestimation between composite endpoint and multi-state analysis in cardiovascular trials

Background

Composite endpoints comprising hospital admissions and death are the primary outcome in many cardiovascular clinical trials. For statistical analysis, a Cox proportional hazards model for the time to first event is commonly applied. There is an ongoing debate on whether multiple episodes per individual should be incorporated into the primary analysis. While the advantages in terms of power are readily apparent, potential biases have been mostly overlooked so far.

Methods

Motivated by a randomized controlled clinical trial in heart failure patients, we use directed acyclic graphs (DAG) to investigate potential sources of bias in treatment effect estimates, depending on whether only the first or multiple episodes are considered. The biases first are explained in simplified examples and then more thoroughly investigated in simulation studies that mimic realistic patterns.

Results

Particularly the Cox model is prone to potentially severe selection bias and direct effect bias, resulting in underestimation when restricting the analysis to first events. We find that both kinds of bias can simultaneously be reduced by adequately incorporating recurrent events into the analysis model. Correspondingly, we point out appropriate proportional hazards-based multi-state models for decreasing bias and increasing power when analyzing multiple-episode composite endpoints in randomized clinical trials.

Conclusions

Incorporating multiple episodes per individual into the primary analysis can reduce the bias of a treatment’s total effect estimate. Our findings will help to move beyond the paradigm of considering first events only for approaches that use more information from the trial and augment interpretability, as has been called for in cardiovascular research.

Yuen C, Rich ML, Evans SR. Totality of outcomes: A different paradigm in assessing interventions for treatment of tuberculosis

Conventional analytic methods used for tuberculosis (TB) outcomes research use standardized outcomes definitions and assess safety and efficacy separately. These methods are subject to important limitations. Conventionally utilized outcome definitions fail to capture important aspects of patients' treatment experience and obscure meaningful differences between patients. Assessing safety and efficacy separately fails to yield an objective risk–benefit comparison to guide clinical practice. We propose to address these issues through an analytic approach based on prioritized outcomes. This approach enables a more comprehensive and integrated assessment of TB interventions. It simultaneously considers a “totality of outcomes”, including clinical benefit, adverse events, and quality of life. These composite outcomes are ranked terms of overall desirability and compared using statistical methods for ordinal outcomes. Here we discuss the application of this approach to TB research, the considerations involved with prioritizing TB treatment outcomes, and the statistical methods involved in comparing prioritized outcomes.

Large sample inference for a win ratio analysis of a composite outcome based on prioritized components

Composite outcomes are common in clinical trials, especially for multiple time-to-event outcomes (endpoints). The standard approach that uses the time to the first outcome event has important limitations. Several alternative approaches have been proposed to compare treatment versus control, including the proportion in favor of treatment and the win ratio. Herein, we construct tests of significance and confidence intervals in the context of composite outcomes based on prioritized components using the large sample distribution of certain multivariate multi-sample U-statistics. This non-parametric approach provides a general inference for both the proportion in favor of treatment and the win ratio, and can be extended to stratified analyses and the comparison of more than two groups. The proposed methods are illustrated with time-to-event outcomes data from a clinical trial.

Heart failure re-admission: measuring the ever shortening gap between repeat heart failure hospitalizations

Many quality-of-care and risk prediction metrics rely on time to first rehospitalization even though heart failure (HF) patients may undergo several repeat hospitalizations. The aim of this study is to compare repeat hospitalization models. Using a population-based cohort of 40,667 patients, we examined both HF and all cause re-hospitalizations using up to five years of follow-up. Two models were examined: the gap-time model which estimates the adjusted time between hospitalizations and a multistate model which considered patients to be in one of four states; community-dwelling, in hospital for HF, in hospital for any reason, or dead. The transition probabilities and times were then modeled using patient characteristics and number of repeat hospitalizations. We found that during the five years of follow-up roughly half of the patients returned for a subsequent hospitalization for each repeat hospitalization. Additionally, we noted that the unadjusted time between hospitalizations was reduced ∼40% between each successive hospitalization. After adjustment each additional hospitalization was associated with a 28 day (95% CI: 22-35) reduction in time spent out of hospital. A similar pattern was seen when considering the four state model. A large proportion of patients had multiple repeat hospitalizations. Extending the gap between hospitalizations should be an important goal of treatment evaluation.

Treatment selections using risk-benefit profiles based on data from comparative randomized clinical trials with multiple endpoints

In a typical randomized clinical study to compare a new treatment with a control, oftentimes each study subject may experience any of several distinct outcomes during the study period, which collectively define the "risk-benefit" profile. To assess the effect of treatment, it is desirable to utilize the entirety of such outcome information. The times to these events, however, may not be observed completely due to, for example, competing risks or administrative censoring. The standard analyses based on the time to the first event, or individual component analyses with respect to each event time, are not ideal. In this paper, we classify each patient's risk-benefit profile, by considering all event times during follow-up, into several clinically meaningful ordinal categories. We first show how to make inferences for the treatment difference in a two-sample setting where categorical data are incomplete due to censoring. We then present a systematic procedure to identify patients who would benefit from a specific treatment using baseline covariate information. To obtain a valid and efficient system for personalized medicine, we utilize a cross-validation method for model building and evaluation and then make inferences using the final selected prediction procedure with an independent data set. The proposal is illustrated with the data from a clinical trial to evaluate a beta-blocker for treating chronic heart failure patients.