Independence estimators for re-randomisation trials in multi-episode settings: a simulation study

Blood eosinophil-guided oral prednisolone for COPD exacerbations in primary care in the UK (STARR2): a non-inferiority, multicentre, double-blind, placebo-controlled, randomised controlled trial

BACKGROUND

Systemic glucocorticoids are recommended for use in chronic obstructive pulmonary disease (COPD) exacerbations; however, there is increased harm associated with their use. We hypothesised that the use of eosinophil biomarker-directed oral prednisolone therapy at the time of an exacerbation of COPD was effective at reducing prednisolone use without affecting adverse outcomes.

METHODS

The studying acute exacerbations and response (STARR2) study was a multicentre, randomised, double-blind, placebo-controlled trial conducted in 14 primary care practices in the UK. We included adults (aged ≥40 years), who were current or former smokers (with at least a 10 pack year smoking history) with a diagnosis of COPD, defined as a post-bronchodilator FEV1/forced vital capacity ratio of less than 0·7 previously recorded by the primary care physician, and a history of at least one exacerbation in the previous 12 months requiring systemic corticosteroids with or without antibiotics. All study staff and participants were masked to study group allocation and to treatment allocation. Participants were randomly assigned (1:1) to blood eosinophil-directed treatment (BET; to receive oral prednisolone 30 mg once daily if eosinophil count was high [≥2%] or placebo if eosinophil count was low [<2%]) or to standard care treatment (ST; to receive prednisolone 30 mg once daily irrespective of the point-of-care eosinophil result). Treatment was prescribed for 14 days and all patients also received antibiotics. The primary outcome was the rate of treatment failure, defined as any need for re-treatment with antibiotics or steroids, hospitalisation for any cause, or death, assessed at 30 days after exacerbation in the modified intention-to-treat population. Participants were eligible for re-randomisation at further exacerbations (with a maximum of four exacerbations per participant). A safety analysis was conducted on all randomly assigned participants. Although designed as a superiority trial, after identification of an error in the randomisation code before data lock the study converted to show non-inferiority. An upper margin of 1·105 for the 95% CI was defined as the non-inferiority margin. This study was registered with EudraCT, 2017-001586-24, and is complete.

FINDINGS

Between Nov 6, 2017, and April 30, 2020, 308 participants were recruited from 14 general practices. 144 exacerbations (73 in the BET group and 71 in the ST group) from 93 participants (mean age 70 years [range 46-84] and mean percent predicted FEV1 60·9% [SD 19·4]; 52 [56%] male and 41 [44%] female; ethnicity data was not collected]) were included in the modified intention-to-treat analysis. There were 14 (19%) treatment failures at 30 days post-exacerbation in the BET group and 23 (32%) in the ST group; we found a large non-significant estimated effect between BET and ST (RR 0·60 [95% CI 0·33-1·04]; p=0·070) in reducing treatment failures after a COPD exacerbation. The non-inferiority analysis supported that BET was non-inferior to ST. Frequency of adverse events were similar between the study groups; glycosuria (2/102 [2%] in BET group and 1/101 [1%] in the ST group) and hospital admission for COPD exacerbation (2/102 [2%] in BET group and 1/101 [1%] in the ST group) were the two most common adverse events in both groups. No deaths occurred in the study.

INTERPRETATION

Blood eosinophil-directed prednisolone therapy at the time of an acute exacerbation of COPD is non-inferior to standard care and can be used to safely reduce systemic glucocorticoid use in clinical practice.

FUNDING

National Institute for Health and Care Research.

Eliminating Ambiguous Treatment Effects Using Estimands

Most reported treatment effects in medical research studies are ambiguously defined, which can lead to misinterpretation of study results. This is because most authors do not attempt to describe what the treatment effect represents, and instead require readers to deduce this based on the reported statistical methods. However, this approach is challenging, because many methods provide counterintuitive results. For example, some methods include data from all patients, yet the resulting treatment effect applies only to a subset of patients, whereas other methods will exclude certain patients while results will apply to everyone. Additionally, some analyses provide estimates pertaining to hypothetical settings in which patients never die or discontinue treatment. Herein we introduce estimands as a solution to the aforementioned problem. An estimand is a clear description of what the treatment effect represents, thus saving readers the necessity of trying to infer this from study methods and potentially getting it wrong. We provide examples of how estimands can remove ambiguity from reported treatment effects and describe their current use in practice. The crux of our argument is that readers should not have to infer what investigators are estimating; they should be told explicitly.

Using re-randomisation designs to increase the efficiency and applicability of retention studies within trials: a case study

BACKGROUND

Poor retention in randomised trials can lead to serious consequences to their validity. Studies within trials (SWATs) are used to identify the most effective interventions to increase retention. Many interventions could be applied at any follow-up time point, but SWATs commonly assess interventions at a single time point, which can reduce efficiency.

METHODS

The re-randomisation design allows participants to be re-enrolled and re-randomised whenever a new retention opportunity occurs (i.e. a new follow-up time point where the intervention could be applied). The main advantages are as follows: (a) it allows the estimation of an average effect across time points, thus increasing generalisability; (b) it can be more efficient than a parallel arm trial due to increased sample size; and (c) it allows subgroup analyses to estimate effectiveness at different time points. We present a case study where the re-randomisation design is used in a SWAT.

RESULTS

In our case study, the host trial is a dental trial with two available follow-up points. The Sticker SWAT tests whether adding the trial logo's sticker to the questionnaire's envelope will result in a higher response rate compared with not adding the sticker. The primary outcome is the response rate to postal questionnaires. The re-randomisation design could double the available sample size compared to a parallel arm trial, resulting in the ability to detect an effect size around 28% smaller.

CONCLUSION

The re-randomisation design can increase the efficiency and generalisability of SWATs for trials with multiple follow-up time points.

Partially clustered designs for clinical trials: Unifying existing designs using consistent terminology

INTRODUCTION

Clinical trial designs based on the assumption of independent observations are well established. Clustered clinical trial designs, where all observational units belong to a cluster and outcomes within clusters are expected to be correlated, have also received considerable attention. However, many clinical trials involve partially clustered data, where only some observational units belong to a cluster. Examples of such trials occur in neonatology, where participants include infants from both singleton and multiple births, and ophthalmology, where one or two eyes per participant may need treatment. Partial clustering can also arise in trials of group-based treatments (e.g. group education or counselling sessions) or treatments administered individually by a discrete number of health care professionals (e.g. surgeons or physical therapists), when this is compared to an unclustered control arm. Trials involving partially clustered data have received limited attention in the literature and the current lack of standardised terminology may be hampering the development and dissemination of methods for designing and analysing these trials.

METHODS AND EXAMPLES

In this article, we present an overarching definition of partially clustered trials, bringing together several existing trial designs including those for group-based treatments, clustering due to facilitator effects and the re-randomisation design. We define and describe four types of partially clustered trial designs, characterised by whether the clustering occurs pre-randomisation or post-randomisation and, in the case of pre-randomisation clustering, by the method of randomisation that is used for the clustered observations (individual randomisation, cluster randomisation or balanced randomisation within clusters). Real life examples are provided to highlight the occurrence of partially clustered trials across a variety of fields. To assess how partially clustered trials are currently reported, we review published reports of partially clustered trials.

DISCUSSION

Our findings demonstrate that the description of these trials is often incomplete and the terminology used to describe the trial designs is inconsistent, restricting the ability to identify these trials in the literature. By adopting the definitions and terminology presented in this article, the reporting of partially clustered trials can be substantially improved, and we present several recommendations for reporting these trial designs in practice. Greater awareness of partially clustered trials will facilitate more methodological research into their design and analysis, ultimately improving the quality of these trials.

Estimands in cluster-randomized trials: choosing analyses that answer the right question

BACKGROUND

Cluster-randomized trials (CRTs) involve randomizing groups of individuals (e.g. hospitals, schools or villages) to different interventions. Various approaches exist for analysing CRTs but there has been little discussion around the treatment effects (estimands) targeted by each.

METHODS

We describe the different estimands that can be addressed through CRTs and demonstrate how choices between different analytic approaches can impact the interpretation of results by fundamentally changing the question being asked, or, equivalently, the target estimand.

RESULTS

CRTs can address either the participant-average treatment effect (the average treatment effect across participants) or the cluster-average treatment effect (the average treatment effect across clusters). These two estimands can differ when participant outcomes or the treatment effect depends on the cluster size (referred to as 'informative cluster size'), which can occur for reasons such as differences in staffing levels or types of participants between small and large clusters. Furthermore, common estimators, such as mixed-effects models or generalized estimating equations with an exchangeable working correlation structure, can produce biased estimates for both the participant-average and cluster-average treatment effects when cluster size is informative. We describe alternative estimators (independence estimating equations and cluster-level analyses) that are unbiased for CRTs even when informative cluster size is present.

CONCLUSION

We conclude that careful specification of the estimand at the outset can ensure that the study question being addressed is clear and relevant, and, in turn, that the selected estimator provides an unbiased estimate of the desired quantity.

Re-randomisation trials in multi-episode settings: Estimands and independence estimators

Often patients may require treatment on multiple occasions. The re-randomisation design can be used in such multi-episode settings, as it allows patients to be re-enrolled and re-randomised for each new treatment episode they experience. We propose a set of estimands that can be used in multi-episode settings, focusing on issues unique to multi-episode settings, namely how each episode should be weighted, how the patient's treatment history in previous episodes should be handled, and whether episode-specific effects or average effects across all episodes should be used. We then propose independence estimators for each estimand, and show the manner in which many re-randomisation trials have been analysed in the past (a simple comparison between all intervention episodes vs. all control episodes) corresponds to a per-episode added-benefit estimand, that is, the average effect of the intervention across all episodes, over and above any benefit conferred from the intervention in previous episodes. We show this estimator is generally unbiased, and describe when other estimators will be unbiased. We conclude that (i) consideration of these estimands can help guide the choice of which analysis method is most appropriate; and (ii) the re-randomisation design with an independence estimator can be a useful approach in multi-episode settings.