Methodological issues in relation to the use of surrogate and composite outcome measures in critical care trials

Abstract

Clinical trials are crucial in the development of efficient healthcare interventions. Choice of outcome measure to estimate the efficacy of an intervention is a critical decision in a study. The National Institute of Health (NIH) definition working group defines ‘clinical endpoint’ as “a characteristic or variable that reflects how a patient feels, functions or survives” and “a surrogate endpoint is a biomarker intended to substitute for a clinical endpoint that should predict clinical benefit or harm or lack of both”. Composite endpoints combine two or more components to form a single outcome, which helps to increase the event rate. Critical care studies have challenges in the analysis of commonly used composite outcome measures. The overarching aim of this PhD was to investigate methodological issues related to the use of surrogate and composite outcome measures in critical care trials. The project comes under the umbrella of outcome measures, and it is divided into two studies. Study 1 examines the variability and validity of surrogate outcome measures and study 2 explores the composite outcome measures in critical care. Three different techniques were used in this thesis: Systematic review, survey, and statistical analysis of secondary data.

The initial literature review discussed different intervention development models, phase II designs and surrogate models. A systematic review of 48 studies published in six journals in the critical care category with the highest impact factor at the time was included in this review. The review aimed to generate a list of surrogate outcome measures reported in critical care studies, determine the variability in the outcome reporting and create a smaller list of potential early phase endpoints for critical care trials. Critical care clinicians and general healthcare researchers were surveyed to gather their views on phase II studies, surrogate outcomes, and standardisation of outcomes. Critical care clinicians and researchers were also asked about their views on the impact of these outcomes in clinical decision making, study progress and overall relevance of the outcomes identified in the systematic review. The critical care researchers and clinicians considered organ failure outcomes as the most influential outcomes followed by the length of stay outcomes, disease severity scores, routinely collected physiology variables and finally biomarkers.
The subsequent statistical analysis investigated the issue of analysed duration outcomes, Sequential Organ Failure Assessment (SOFA) score, PaCO2/FiO2ratio (PF ratio) and Oxygenation Index (OI). A joint modelling approach for analysing mechanical ventilation in the presence of a terminal event was explored. The association between the outcome measures such as the score, PaCO2/FiO2ratio (PF ratio), Oxygenation Index (OI) and the clinical endpoint, mortality was investigated using joint modelling. Joint modelling estimates were compared with cox regression estimates. Mediation analysis was used to estimate the direct and indirect effects. The estimates based on mediation analysis were compared was joint modelling estimates.
The study proposes the following core outcomes for early phase MV studies
1. Mortality
2. Duration of MV
3. Length of stay in ICU
4. Length of stay in hospital
5. SOFA
6. OI
7. PF ratio
The study on composite outcome measures included a literature review of composite outcomes in general and the utility of the VFD score as an outcome in critical care studies. The first approach tested whether any of the count data models such as Poisson, Zero Inflated Poisson (ZIP) and Logit-Poisson hurdle model provided a better fit for the VFD score compared to the traditional t-test. The win-ratio method is widely used as an approach to analysing composites with components arranged in hierarchical order. The usefulness of different combinations of mortality and MV outcomes was investigated. The use of an additional outcome, SOFA as a tie-breaker for win-ratio analysis was explored. A two-part hurdle was a promising approach to analysing VFD. Future studies could investigate the sample size requirement. Win-ratio analysis was easy to understand and simple to implement. For early phase studies, mortality, and duration of MV was recommended as outcomes to categorise patients into winner and loser.

Thesis embargoed until 31 July 2025.

Date of Award	Jul 2022
Original language	English
Awarding Institution	Queen's University Belfast
Sponsors	Northern Ireland Clinical Trials Unit
Supervisor	Danny McAuley (Supervisor), Cliona McDowell (Supervisor), Bronagh Blackwood (Supervisor) & Mike Clarke (Supervisor)