Composite Endpoints in CKD Trials Show Bias Driven by Event Frequency, Not Complexity, for SGLT2i, GLP-1 RA, MRA
Background
Chronic kidney disease (CKD) is a major global health challenge, often co-occurring with cardiovascular disease, leading to high morbidity and mortality. Clinical trials in CKD frequently employ composite endpoints to boost statistical power and efficiency, especially given the long follow-up and relatively low event rates for individual outcomes like kidney failure or cardiovascular death. However, these composites can obscure the true clinical impact if less meaningful components disproportionately drive the overall effect. Understanding this potential bias is crucial for accurately interpreting the benefits of emerging cardiorenal protective therapies like SGLT2 inhibitors, GLP-1 receptor agonists, and mineralocorticoid receptor antagonists.
Study Design
Researchers conducted a meta-epidemiological analysis of eight randomized controlled trials (RCTs) involving 38 composite endpoints that evaluated sodium-glucose cotransporter 2 inhibitors, glucagon-like peptide-1 receptor agonists, and non-steroidal mineralocorticoid receptor antagonists in CKD patients. The study defined the Bias Attributable to Composite Outcome (BACO) index as the ratio of the log-hazard ratio (HR) for the composite endpoint to that of a pre-specified reference outcome (kidney failure or cardiovascular death). Variance was estimated using the delta method, and determinants of BACO variability were analyzed using inverse-variance weighted mixed-effects meta-regression.
Results
The analysis revealed that higher reference-event rates were significantly associated with higher BACO values, both overall (β: 0.06, 95% CI: 0.02; 0.10) and specifically in kidney failure-referenced analyses (β: 0.07, 95% CI: 0.02; 0.12). Stronger composite treatment effects also correlated with higher BACO (β: -1.07, 95% CI: -1.84; -0.30), indicating that a larger treatment effect on the composite might be driven by components that are more frequent or more responsive to treatment. The number of components within a composite endpoint and the duration of follow-up showed no significant association with BACO. In models referenced to cardiovascular death, BACO was associated with trial size (β: 0.12 per 1000 participants), mean age (β: -0.04 per 10 years), and female proportion (β: 0.09 per 10% increase).
Key Findings
- Higher reference-event rates correlated with higher BACO values overall (β: 0.06, 95% CI: 0.02; 0.10).
- Stronger composite treatment effects were associated with higher BACO (β: -1.07, 95% CI: -1.84; -0.30).
- Number of components and follow-up duration showed no significant association with BACO.
- In cardiovascular death-referenced models, BACO was linked to trial size (β: 0.12 per 1000 participants), mean age (β: -0.04 per 10 years), and female proportion (β: 0.09 per 10% increase).
Why It Matters
This study highlights a critical consideration for researchers and clinicians interpreting or designing trials for CKD and cardiovascular disease. Composite endpoints may not reliably reflect treatment effects if the clinically most important outcomes are infrequent or less responsive to the intervention. This implies that simply adding more components to a composite endpoint does not necessarily improve its clinical relevance or reduce bias. Future trial designs should prioritize clinically aligned endpoint strategies, ensuring that the primary components of composite outcomes are both frequent enough and sufficiently sensitive to treatment effects to accurately represent the true patient benefit. This could lead to more transparent and impactful reporting of drug efficacy for cardiorenal protection.
ckd
cardiovascular-disease
composite-endpoints
meta-epidemiological
sglt2-inhibitors
glp-1-agonists