Clinical Scorecard: Addressing Time-Dependent Bias in Observational Studies of ECPR

At a Glance

Key Highlights

• Early ROSC among CCPR patients can bias comparisons against ECPR, while immortal time before cannulation can bias them in favor of ECPR.
• ECPR decision timing is often difficult to identify retrospectively, and registries may not record when CCPR becomes a bridge to ECPR.
• Few clinically comparable CCPR patients may remain after 20–30 minutes of advanced life support, limiting causal inference at prolonged low-flow durations.
• Time-dependent regression, propensity score methods, landmark analyses, and target trial emulation may partly address bias, but each has important limitations.
• The authors prioritize standardized registry timestamps, followed by prospective multicenter collection of time-varying data.

Guideline-Based Recommendations

Diagnosis

• The authors discuss assessing whether time zero is synchronized and whether low-flow duration and OHCA severity are addressed when interpreting observational ECPR comparisons.
• Distinguish resuscitation time bias from immortal time bias when evaluating treatment-group definitions and survival comparisons.

Management

• The authors identify standardized timestamps for the ECPR decision, cannulation, ECMO flow, and resuscitation termination as a registry priority.
• The authors propose prospective multicenter observational studies collecting time-varying data to quantify these biases.
• The article notes that statistical adjustment may partly reduce bias but cannot resolve structural limitations.

Monitoring & Follow-up

• Use standardized definitions and timestamps to specify time zero accurately in registries.
• Assess overlap in low-flow duration and OHCA clinical severity when applying time-dependent regression or propensity score-based approaches.

Risks

• Early ROSC in the CCPR group may make those patients less comparable with patients experiencing prolonged OHCA.
• Assigning survival time before ECPR pump initiation to the CCPR group can produce confounding attributable to study design.
• Landmark analyses may exclude patients whose CCPR ended before the threshold; lowering the threshold may reintroduce resuscitation time bias.
• At prolonged low-flow durations, continued CCPR may not be an equivalent therapeutic alternative to ECPR, limiting experimental positivity and causal inference.

Patient & Prescribing Data

ECPR is described as a second-line option for selected patients with refractory OHCA when trained teams and appropriately equipped care systems are available.

The article discusses observational comparisons of ECPR with CCPR, not prescribing. Reported intervals from the ECPR decision to pump initiation are not specified; the source reports logistical delays of 20 minutes in INCEPTION and 12 minutes in ARREST and Belohlavek et al.

Clinical Best Practices

• The authors discuss specifying and synchronizing time zero when designing or interpreting observational ECPR comparisons.
• The authors discuss measuring low-flow duration and OHCA clinical severity before exposure when applying adjustment or matching methods.
• The authors identify standardized recording of ECPR decision timing and key resuscitation events as a registry priority.
• The authors advise cautious interpretation when analyses do not synchronize time zero or insufficiently address low-flow duration.
• The article describes validity, relevance, sample-size, and confounding challenges for instrumental-variable and interrupted time-series approaches.

Related Resources & Content