Mar 31, 2026
Bias in the composite outcomes of kidney-cardio protective randomized controlled trials in chronic kidney disease: a meta-epidemiological study
- Ioannis Bellos1,
- Vassiliki Benetou1
- 1Department of Hygiene, Epidemiology and Medical Statistics, National and Kapodistrian University of Athens, Medical School, Athens, Greece
Protocol Citation: Ioannis Bellos, Vassiliki Benetou 2026. Bias in the composite outcomes of kidney-cardio protective randomized controlled trials in chronic kidney disease: a meta-epidemiological study. protocols.io https://dx.doi.org/10.17504/protocols.io.yxmvm81wog3p/v1
License: This is an open access protocol distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Protocol status: Working
We use this protocol and it's working
Created: March 31, 2026
Last Modified: March 31, 2026
Protocol Integer ID: 314184
Keywords: trials in chronic kidney disease, treatment effect on the composite outcome, epidemiological study chronic kidney disease, key limitation of composite outcome, composite outcome, bias in the composite outcome, chronic kidney disease, bias attributable to composite outcome, evaluating kidney, kidney disease progression, stage kidney disease, ckd trial, kidney failure, overall treatment effect, kidney, cardio protective therapy, controlled trial, meaningful outcomes such as end, cardiovascular event, cardiovascular disease, cardiovascular domain, meaningful outcome
Abstract
Chronic kidney disease (CKD) is characterized by a high burden of morbidity and mortality, associated with kidney failure and cardiovascular disease. Randomized controlled trials (RCTs) evaluating kidney–cardio protective therapies increasingly rely on composite primary endpoints that combine kidney disease progression, cardiovascular events, and death. While this strategy improves statistical efficiency and feasibility, it introduces important challenges in interpretation, as the individual components of these composites differ substantially in frequency, severity, and clinical relevance. A key limitation of composite outcomes is that the overall treatment effect may be disproportionately influenced by more frequent or less clinically consequential events, potentially obscuring the effect on the most meaningful outcomes such as end-stage kidney disease (ESKD) or death. The Bias Attributable to Composite Outcome (BACO) index has been proposed to quantify this phenomenon by comparing the treatment effect on the composite outcome with that on a key component. In CKD trials, where composite endpoints frequently span both kidney and cardiovascular domains, a clinically grounded application of BACO is required to ensure that interpretation reflects the dominant disease process and the most relevant hard outcomes.
Troubleshooting
Background Chronic kidney disease (CKD) is characterized by a high burden of morbidity and mortality, associated with kidney failure and cardiovascular disease. Randomized controlled trials (RCTs) evaluating kidney–cardio protective therapies increasingly rely on composite primary endpoints that combine kidney disease progression, cardiovascular events, and death. While this strategy improves statistical efficiency and feasibility, it introduces important challenges in interpretation, as the individual components of these composites differ substantially in frequency, severity, and clinical relevance. A key limitation of composite outcomes is that the overall treatment effect may be disproportionately influenced by more frequent or less clinically consequential events, potentially obscuring the effect on the most meaningful outcomes such as end-stage kidney disease (ESKD) or death. The Bias Attributable to Composite Outcome (BACO) index has been proposed to quantify this phenomenon by comparing the treatment effect on the composite outcome with that on a key component. In CKD trials, where composite endpoints frequently span both kidney and cardiovascular domains, a clinically grounded application of BACO is required to ensure that interpretation reflects the dominant disease process and the most relevant hard outcomes.
Objectives The objective of this meta-epidemiological study is to evaluate whether treatment effects reported for composite primary outcomes in randomized trials of kidney–cardio protective therapies in CKD are consistent with treatment effects observed for clinically meaningful component outcomes.
Eligibility criteria The study will include RCTs evaluating the efficacy of novel kidney–cardio protective therapies in patients with chronic kidney disease. The interventions of interest will consist of SGLT2 (sodium glucose cotransporter 2) inhibitors, GLP1 (glucagon-like peptide-1) receptor agonists and non-steroidal mineralocorticoid receptor antagonists (ns-MRA). The interventions will be compared to placebo. Eligible studies should report at least one composite endpoint, a hazard ratio (HR) with 95% confidence intervals (CI) for the composite endpoint and HRs with 95% CIs for at least one clinically relevant component outcome, including kidney failure, cardiovascular death or all-cause death. Studies that do not provide sufficient data to estimate treatment effects for both the composite outcome and a component outcome will be excluded. Composite primary endpoints will be classified as: kidney-directed, when primarily composed of kidney disease progression outcomes, cardiovascular-directed, when primarily composed of cardiovascular outcomes , mixed, when both kidney and cardiovascular components contribute substantially to the endpoint
Search strategy The following databases will be systematically searched: PubMed, Scopus, Web of Science and Clinicaltrials.gov. All databases will be searched from their inception. The search algorithm will include a combination of MeSH (Medical Subject Headings) terms and key-words of the interventions of interest (both class and drug-specific). The literature search will be performed by two researchers independently, resolving any discrepancies through their consensus.
BACO calculation For each trial, the BACO index will be calculated as: BACO = Log (HRcomposite) / Log (HRreference), where HRcomposite is the hazard ratio for the composite outcome and HRreference is the hazard ratio for the reference, key component, outcome. Log-transformed hazard ratios will be used for all calculations. Standard errors will be derived from reported confidence intervals: SE = [Log (𝑈) – Log (𝐿)]/ 2 x 1.96, where 𝐿 and 𝑈 represent the lower and upper limits of the 95% CI. The primary analysis will focus on BACO point estimates. Trials will be categorized as: overestimation, when BACO > 1, attenuation, when 0 < BACO < 1, and inversion, when BACO < 0. Estimates will be considered unstable when the reference hazard ratio is close to 1 or when reference events are sparse. The BACO index will be interpreted as an indicator of the extent to which the treatment effect on a composite endpoint reflects the effect on a clinically meaningful component outcome. Particular attention will be given to discrepancies between composite-driven and component-driven treatment effects.
Data extraction The following data will be extracted for trial description: trial name or registration number, year of publication, geographic region, intervention, sample size, blinding status, and patient characteristics (age, sex, BMI, estimated glomerular filtration rate [eGFR], albuminuria, diabetes status, history of cardiovascular disease). Additional trial characteristics will include follow-up duration, heart failure enrichment, albuminuria enrichment, and early trial termination. For the main analysis, the definition of composite endpoints will be recorded, along with the hazard ratio (HR) and 95% confidence interval (CI) for both the composite outcome and the relevant component outcome (kidney failure, cardiovascular death, or all-cause death). The following methodological characteristics of composite endpoints will also be extracted: number of components in the composite, inclusion of softer components such as hospitalization events or eGFR decline thresholds, whether the reference outcome is included within the composite, total number of composite events, and total number of reference events.