Composite Endpoints in Clinical Trials

Composite endpoints can encode multiple pieces of information and are increasingly adopted in clinical trials. Advocacy for using composite endpoints began decades ago in cardiovascular trials, leading to incorporation of patient-oriented outcomes and consideration of a hierarchical ranking system. The use of composite endpoints in coronavirus disease (COVID-19) trials has evolved similarly. We conducted a literature review to investigate the use of composite endpoints in acute heart failure and COVID-19 clinical trials. The results showed more frequent use of patient-oriented outcomes and ordinal composite endpoints in COVID-19 trials, which might be driven by global consensus on a set of common outcome measures.

Composite Endpoints in Clinical Trials

This article discusses some important issues that may arise in the current usage of composite endpoints as primary endpoints for demonstrating the efficacy of new drugs in clinical trials. The discussion focuses on time-to-event composite endpoints. Issues discussed include validity of a composite endpoint, the often lack of follow-up of patients beyond first event, the analysis of a composite endpoint, its sub-composite and individual component endpoints and their interpretation. Actual published examples in the literature are used to illustrate some of these problems. It is recommended that a clinical trial using a composite endpoint as the primary endpoint should be designed to include patient follow-up beyond the first event if possible. For data collected from such trials, basic formats for tabular presentation of trial data and for results of analysis of the composite endpoint, its sub-composite and individual component endpoints are proposed for transparency and ease of interpretation.

Reply: Composite endpoints in clinical trials - simplicity or perfection?

Two issues on clinical trials with multiple endpoints were surveyed: (1) the terminology of multiple endpoints, relationship between rare events and endpoints, and differences in multiplicity adjustment between regions; and (2) the current practice on multiplicity adjustment and sample size calculation. This article provides a summary of the results of a survey on the first issue. The survey was conducted among 63 members of the Japan Pharmaceutical Manufacturers Association from October to November 2017. Thirty-five companies based in Japan and 12 companies based in other countries, 47 companies in total, responded to the survey. The terms co-primary endpoints, secondary endpoint, and composite endpoint were used in a variety of ways. An endpoint for a clinically most important event that is expected to occur rarely differed between regions. Although the Pharmaceuticals and Medical Devices Agency did not demand multiplicity adjustment, it was considered in clinical trials with multiple endpoints for approval in Japan. The use of terminology differed from the definition in the Food and Drug Administration guidance and the European Medicines Agency guideline. There remain challenges on a clinically most important event that is expected to occur rarely and multiplicity adjustment in clinical trials with multiple endpoints.

Composite endpoints in clinical trials: the “win ratio”

To The Editors: The US Food and Drug Administration (FDA) has held several meetings discussing clinical trial design for studies of acute otitis media (AOM), 1 including discussions of appropriate endpoints for such trials. The results of the study by Piglansky et al. 2 are based on bacteriologic outcomes as the primary endpoint. The premise of utilizing a bacteriologic outcome as the primary endpoint is based on an understanding of the correlation between clinical and bacteriologic outcomes. Although antimicrobials have a direct effect on microorganisms, this effect should ultimately translate into diminution of the symptoms of the infection for which patients seek treatment. In this study the investigators evaluated 50 children with AOM in a double tympanocentesis study using high dose amoxicillin (80 mg/kg/day 3 times a day for 10 days). The bacteriologic success rate at Day 4 to 6 was 72% (36 of 50), whereas the clinical success rate at Day 4 of 6 was 94% (47 of 50). This translates into a number of children designated as bacteriologic failures on the second tympanocentesis who were clinically cured. The paper does not provide a patient-by-patient listing relating each patient’s bacterial and clinical outcomes; however, there were three patients described as clinical failures who were also bacteriologic failures. Using these numbers we calculated the correlation between the bacteriologic and clinical outcomes (Table 1). 3 The kappa coefficient for the study was 0.28 with a standard deviation of 0.13, showing only fair agreement between the two endpoints.TABLE 1: Comparison of bacteriologic and clinical outcomes in three double tympanocentesis acute otitis media studiesAt a recent FDA meeting on clinical trial study considerations for AOM, 1 several presentations advocated the use of bacteriologic endpoints in double tympanocentesis studies as the primary endpoint. However, the kappa coefficient for the two studies 4, 5 discussed at the advisory meeting (kappa = 0.34 and kappa = 0.35) also showed less than substantial 3 correlation between bacteriologic and clinical endpoints. The cited studies supported a “correlation”4, 5 between children who clinically failed treatment and those who were culture-positive on the second tympanocentesis. To evaluate a true correlation, however, one must also look at the number of children who were clinical successes who remained culture-positive. In both studies as well as the current study by Piglansky et al., 2 substantial numbers of children were clinically cured despite positive cultures on the second tympanocentesis. The chi square test and the associated P value presented in these studies show an association of the two variables but do not provide a quantitative measure of the degree of the relationship between clinical and bacteriologic outcomes. The significant association of the two outcomes is driven by the high sensitivity (number of clinical and bacteriologic success patients/total bacteriologic success patients) of clinical failure for predicting bacteriologic failure. However, the specificity of bacteriologic failure for predicting clinical failure (number of clinical and bacteriologic failures/total bacteriologic failure patients) is too low for use in a clinical trial setting when attempting to differentiate outcomes between treatment arms (Table 1). Although high sensitivity is desirable in both clinical practice and clinical trials, high specificity is also important in clinical trials attempting to assess antimicrobial efficacy. Double tympanocentesis studies in clinical trials for AOM are important for assessment of the bacteriology and the natural history of the disease. However, the less than substantial correlation between bacteriologic and clinical outcomes raises significant questions about using bacteriologic outcomes as the sole measure of drug efficacy. Authors have claimed that clinical endpoints do not sufficiently discriminate between the efficacies of various therapies in AOM. 6 However, it may be that clinical resolution measured at a fixed time point well beyond the expected time of natural resolution of the disease (Day 10 or Day 28) may not provide the most useful information. Time-to-resolution of symptoms, used in other indications with high spontaneous cure rates, 7 may be more meaningful clinical endpoints in AOM clinical trials. Perhaps the intrapatient correlation between the two endpoints would be improved with use of such clinical endpoints. Rosemary Johann-Liang, M.D. Jyoti Zalkikar, Ph.D. John H. Powers, M.D.

assumption is often untenable, necessitating a conceptual framework with greater specificity.

Incorporating Data from Multiple Endpoints in the Analysis of Clinical Trials: Example from RSV Vaccines.

Cardiovascular mortality has decreased over the past 5 decades, making it increasingly difficult to demonstrate significant benefits of new therapies in randomized clinical trials. We sought to determine whether the use of composite end points in major cardiovascular trials has changed over time and examine temporal trends in the clinical importance of individual components of these composite end points. Using a validated search strategy, we searched MEDLINE for randomized trials of therapies for primary or secondary cardiovascular prevention published in New England Journal of Medicine, The Lancet, and the Journal of the American Medical Association between 1986 and 2015. We abstracted and categorized study population demographics, type of intervention, and primary and secondary outcomes. Composite end point components were ranked according to importance (minor, moderate, major, critical, and death) and temporal trends analyzed. In total, 604 of 2607 trials retrieved met inclusion criteria. Use of composite end points increased significantly over time from 18.8% between 1986 and 1990 to 83.0% between 2011 and 2015 (P<0.001). The number of components in the primary end point also increased significantly (median 1 in 1986-1990, median 3 in 2011-2015; P<0.001). Contemporary trials were more likely to include end points of lesser importance to patients (minor 3.1% and moderate 6.3% in 1986-1990, minor 4.5% and moderate 44.6% in 2011-2015; P<0.001). Use of death as the sole primary end point declined significantly over time (53.1% in 1986-1990, 17.9% in 2011-2015; P<0.001). Contemporary cardiovascular randomized clinical trials are more likely to use primary composite end points that contain a larger number of components. Furthermore, these composite end points have increasingly incorporated components of lesser clinical importance. Clinicians and policymakers interpreting results of randomized trials should recognize that the significance of individual components in a composite end point is often heterogeneous.

Using GFR, Albuminuria, and Their Changes in Clinical Trials and Clinical Care

Multiplicity 2: Multiple treatments and multiple outcomes

Consider the choice of outcome for overall treatment benefit in a clinical trial which measures the first time to each of several clinical events. We describe several new variants of the win ratio that incorporate the time spent in each clinical state over the common follow-up, where clinical state means the worst clinical event that has occurred by that time. One version allows restriction so that death during follow-up is most important, while time spent in other clinical states is still accounted for. Three other variants are described; one is based on the average pairwise win time, one creates a continuous outcome for each participant based on expected win times against a reference distribution and another that uses the estimated distributions of clinical state to compare the treatment arms. Finally, a combination testing approach is described to give robust power for detecting treatment benefit across a broad range of alternatives. These new methods are designed to be closer to the overall treatment benefit/harm from a patient's perspective, compared to the ordinary win ratio. The new methods are compared to the composite event approach and the ordinary win ratio. Simulations show that when overall treatment benefit on death is substantial, the variants based on either the participants' expected win times (EWTs) against a reference distribution or estimated clinical state distributions have substantially higher power than either the pairwise comparison or composite event methods. The methods are illustrated by re-analysis of the trial heart failure: a controlled trial investigating outcomes of exercise training.

Novel approaches to composite endpoints in clinical trials.

79P Comparison of composite endpoints in clinical trials