Reporting of Withdrawal of Life-Sustaining Treatment in Acute Brain Injury Trials

• In ICU
• Tue, 21 Apr 2026

There is an under-recognised methodological challenge in randomised clinical trials (RCTs) involving patients with acute brain injury (ABI): the impact of withdrawal of life-sustaining treatment (WLST) on trial validity and interpretation. WLST represents a significant and often inadequately reported source of bias that can distort estimates of treatment effects, and there is a need for systematic reporting and analytical consideration of this factor in future research.

In ABI trials, the primary goal is typically survival with a neurological outcome that aligns with patient preferences. RCTs frequently rely on neurological endpoints to evaluate the effectiveness of interventions. For such trials to support valid causal conclusions, well-established threats to validity, such as protocol deviations, loss to follow-up, and inconsistent outcome measurement, must be transparently addressed. However, WLST is a less familiar but equally critical threat to causal inference.

WLST is highly prevalent in this patient population, with an estimated 60–70% of deaths in critically ill ABI patients occurring after a decision to withdraw life-sustaining therapies. Because nearly all patients undergoing WLST subsequently die, these decisions have a decisive influence on observed outcomes. WLST practices vary substantially across clinicians, institutions, and regions. Decisions are often guided by neuroprognostication, which attempts to predict long-term neurological outcomes. However, prognostic tools are inherently imperfect, and early predictions, often made during acute phases of illness, are particularly uncertain. Widely used tools such as the Glasgow Coma Scale were not designed for long-term prognostication.

In addition to clinical indicators, WLST decisions are influenced by subjective and contextual factors, including clinician judgement, cognitive biases, institutional culture, communication framing, resource availability, and even health insurance status. These influences persist within RCTs, raising the possibility that differences in outcomes between study groups may reflect variations in care decisions rather than true treatment effects.

The central methodological concern arises from “misclassification” of outcomes. In cases where WLST is implemented, the patient’s “natural” or counterfactual outcome (what would have occurred without withdrawal) is unobservable. If WLST is performed in patients who might otherwise have survived with a favourable neurological outcome, their death represents a misclassified outcome. Such misclassification may occur due to overly pessimistic prognostication or non-clinical influences on decision-making. Evidence suggests that this is not a rare phenomenon.

Randomisation does not protect against this bias because WLST decisions occur after treatment allocation. This means that even well-designed RCTs are vulnerable to distortion. The consequences can be substantial: WLST may alter the apparent effectiveness of an intervention, potentially exaggerating benefit, masking harm, or even reversing the perceived direction of effect. For instance, if clinicians believe a treatment is beneficial and are therefore less likely to withdraw care in that group, improved outcomes may be observed independently of the intervention’s true efficacy. On the other hand, if a treatment is harmful, reduced WLST in that group could obscure its negative effects.

Such biases can arise even in blinded trials and may be exacerbated when WLST decisions are influenced by non-neurological factors, such as comorbidities or perceived quality of life. These complexities make it difficult to interpret trial results when WLST is not adequately accounted for.

Despite growing recognition of this issue, particularly in cardiac arrest research, reporting of WLST in ABI trials remains inconsistent and often incomplete. This lack of transparency hampers the ability to assess whether reported treatment effects are valid or biased. Given the implications for clinical practice, policy-making, and guideline development, this represents a serious concern.

To address this gap, future trials should adopt minimum reporting standards for WLST. These include documenting the frequency of WLST events, the timing of decisions relative to injury and randomisation, and the rationale underlying these decisions, including both prognostic assessments and patient preferences where available. Trials should also conduct sensitivity analyses to explore how WLST may have influenced outcomes, using methods such as outcome imputation or inverse probability censoring.

Beyond descriptive reporting, advanced statistical approaches should be used to account for WLST-related bias. Techniques such as competing risks models and counterfactual simulations can help estimate the range of possible treatment effects under different assumptions. While these methods cannot eliminate uncertainty, they can provide more robust and transparent interpretations of trial findings.

There is also an ethical dimension to WLST. Decisions about life-sustaining treatment are fundamental to patient autonomy and inherently involve subjective judgement. Because these decisions cannot be standardised, their influence on trial outcomes must instead be acknowledged and managed through careful design, analysis, and reporting.

WLST is a critical yet underappreciated determinant of outcomes in ABI trials. Its potential to introduce bias underscores the need for systematic data collection, transparent reporting, and appropriate analytical strategies. Recognising WLST as a key variable is essential for ensuring that trial results are interpretable, reliable, and ethically sound. Only through these measures can the neurocritical care community strengthen causal inference and improve the evidence base guiding treatment decisions in acute brain injury.

Source: AJRCCM
Image Credit: iStock