Dataset Information

Estimation of the optimal surrogate based on a randomized trial.

ABSTRACT: A common scientific problem is to determine a surrogate outcome for a long-term outcome so that future randomized studies can restrict themselves to only collecting the surrogate outcome. We consider the setting that we observe n independent and identically distributed observations of a random variable consisting of baseline covariates, a treatment, a vector of candidate surrogate outcomes at an intermediate time point, and the final outcome of interest at a final time point. We assume the treatment is randomized, conditional on the baseline covariates. The goal is to use these data to learn a most-promising surrogate for use in future trials for inference about a mean contrast treatment effect on the final outcome. We define an optimal surrogate for the current study as the function of the data generating distribution collected by the intermediate time point that satisfies the Prentice definition of a valid surrogate endpoint and that optimally predicts the final outcome: this optimal surrogate is an unknown parameter. We show that this optimal surrogate is a conditional mean and present super-learner and targeted super-learner based estimators, whose predicted outcomes are used as the surrogate in applications. We demonstrate a number of desirable properties of this optimal surrogate and its estimators, and study the methodology in simulations and an application to dengue vaccine efficacy trials.

SUBMITTER: Price BL

PROVIDER: S-EPMC6393111 | biostudies-literature | 2018 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Estimation of the optimal surrogate based on a randomized trial.

Price Brenda L BL Gilbert Peter B PB van der Laan Mark J MJ

Biometrics 20180427 4

A common scientific problem is to determine a surrogate outcome for a long-term outcome so that future randomized studies can restrict themselves to only collecting the surrogate outcome. We consider the setting that we observe n independent and identically distributed observations of a random variable consisting of baseline covariates, a treatment, a vector of candidate surrogate outcomes at an intermediate time point, and the final outcome of interest at a final time point. We assume the treat ...[more]

PMID: 29701875

Similar Datasets

Project description:BackgroundRobust identification of surrogate endpoints can help accelerate the development of pharmacotherapies for diseases traditionally evaluated using true endpoints associated with prolonged follow-up. The meta-analysis-based surrogate endpoint evaluation (SEE) integrates data from multiple, usually smaller, trials to statistically confirm a surrogate endpoint as a robust proxy for the true endpoint. To test the applicability of SEE when only a single, larger trial is available, we analysed the cardiovascular (CV) survival endpoint from the large multinational trial LEADER (9340 subjects) that confirmed the CV safety of a diabetes drug (liraglutide). We evaluated if using country as a trial unit adequately facilitated the meta-analysis and calculation of R2 by country group.MethodsData were grouped by country, ensuring at least 30 CV deaths (497 in total) in each of the nine resulting by-country groups. In a two-step SEE on the grouped dataset, we first fitted the group-specific Cox proportional hazard models; next, on the trial-level, we regressed the estimated hazard ratio (HR; liraglutide vs placebo) of the true endpoints (CV death: 497 events, or all-cause death: 828 events) on the HR of the surrogate endpoint (major CV adverse event [MACE]: 1302 events) and derived the group-specific R2 and its 95% confidence interval (CI).ResultsGroup-level surrogacy of MACE was supported for CV death but not for all-cause death, with [Formula: see text] values of 0.85 [0.63;1.00]95% CI and 0.23 [0.00;0.67]95% CI, respectively. Sensitivity analyses using different grouping approaches (e.g. grouping by region) corroborated the robustness of the conclusions as well as the appropriateness of the data-grouping approaches.ConclusionsWe derived a specific grouping approach to successfully apply SEE on data from a single trial. This may allow for the statistically robust identification and validation of surrogate endpoints based on the abundance of large monolithic outcome trials conducted as part of drug development programmes in, for example, diabetes.

Project description:Rationale: Breakdowns in clinician-family communication in intensive care units (ICUs) are common, yet there are no easily scaled interventions to prevent this problem.Objectives: To assess the feasibility, usability, acceptability, and perceived effectiveness of a communication intervention that pairs proactive family meetings with an interactive, web-based tool to help surrogates prepare for clinician-family meetings.Methods: We conducted a two-arm, single-blind, patient-level randomized trial comparing the Family Support Tool with enhanced usual care in two ICUs in a tertiary-care hospital. Eligible participants included surrogates of incapacitated patients judged by their physicians to have ≥40% risk of death or severe long-term functional impairment. The intervention group received unlimited tool access, with prompts to complete specific content upon enrollment and before two scheduled family meetings. Before family meetings, research staff shared with clinicians a one-page summary of surrogates' main questions, prognostic expectations, beliefs about the patient's values, and attitudes about goals of care. The comparator group received usual care enhanced with scheduled family meetings. Feasibility outcomes included the proportion of participants who accessed the tool before the first family meeting, mean number of logins, and average tool engagement time. We assessed tool usability with the System Usability Scale, assessed tool acceptability and perceived effectiveness with internally developed questionnaires, and assessed quality of communication and shared decision-making using the Quality of Communication questionnaire.Results: Of 182 screened patients, 77 were eligible. We enrolled 52 (67.5%) patients and their primary surrogate. Ninety-six percent of intervention surrogates (24/25) accessed the tool before the first family meeting (mean engagement time, 62 min ± 27.7) and logged in 4.2 times (±2.1) on average throughout the hospitalization. Surrogates reported that the tool was highly usable (mean, 82.4/100), acceptable (mean, 4.5/5 ± 0.9), and effective (mean, 4.4/5 ± 0.2). Compared with the control group, surrogates who used the tool reported higher overall quality of communication (mean, 8.9/10 ± 1.6 vs. 8.0/10 ± 2.4) and higher quality in shared decision-making (mean, 8.7/10 ± 1.5 vs. 8.0/10 ± 2.4), but the difference did not reach statistical significance.Conclusions: It is feasible to deploy an interactive web-based tool to support communication and shared decision-making for surrogates in ICUs. Surrogates and clinicians rated the tool as highly usable, acceptable, and effective.

Dataset Information

Estimation of the optimal surrogate based on a randomized trial.

Publications

Estimation of the optimal surrogate based on a randomized trial.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets