Validity of surrogate endpoints assessing central venous catheter-related infection: evidence from individual- and study-level analyses
Résumé
Objectives: The prevention of catheter-related bloodstream infection (CRBSI) has been an area of intense research, but the heterogeneity of endpoints used to define catheter infection makes the interpretation of randomized controlled trials (RCTs) problematic. The aim of this study was to determine the validity of different endpoints for central venous catheter infections.
Data sources: (a) Individual-catheter data were collected from 9428 catheters from four large RCTs; (b) study-level data from 70 RCTs were identified with a systematic search. Eligible studies were RCTs published between January 1987 and October 2018 investigating various interventions to reduce infections from short-term central venous catheters or short-term dialysis catheters. For each RCT the prevalence rates of CRBSI, quantitative catheter tip colonization, catheter-associated infection (CAI) and central line-associated bloodstream infection (CLABSI) were extracted for each randomized study arm.
Methods: CRBSI was used as the gold-standard endpoint, for which colonization, CAI and CLABSI were evaluated as surrogate endpoints. Surrogate validity was assessed as (1) the individual partial coefficient of determination (individual-pR2) using individual catheter data; (2) the coefficient of determination (study-R2) from mixed-effect models regressing the therapeutic effect size of the surrogates on the effect size of CRBSI, using study-level data.
Results: Colonization showed poor agreement with CRBSI at the individual-patient level (pR2 = 0.33 95% CI 0.28-0.38) and poor capture at the study level (R2 = 0.42, 95% CI 0.21-0.58). CAI showed good agreement with CRBSI at the individual-patient level (pR2 = 0.80, 95% CI 0.76-0.83) and moderate capture at the study level (R2 = 0.71, 95% CI 0.51-0.85). CLABSI showed poor agreement with CRBSI at the individual patient level (pR2 = 0.34, 95% CI 0.23-0.46) and poor capture at the study level (R2 = 0.28, 95% CI 0.07-0.76).
Conclusions: CAI is a moderate to good surrogate endpoint for CRBSI. Colonization and CLABSI do not reliably reflect treatment effects on CRBSI and are consequently more suitable for surveillance than for clinical effectiveness research.