BOX - Fever Control

Foundational Trials

BOX - Fever Control

Hassager C, Schmidt H, Møller JE, Grand J, Mølstrøm S, Beske RP, et al. Duration of device-based fever prevention after cardiac arrest. N Engl J Med. 2023;388(10):888-897

Image: Shutterstock / ChaNaWiT

Publication

Title: Duration of Device-Based Fever Prevention after Cardiac Arrest
Acronym: BOX (platform trial) — Fever Prevention / Fever Control
Year: 2023
Journal published in: New England Journal of Medicine
Citation: Hassager C, Schmidt H, Møller JE, Grand J, Mølstrøm S, Beske RP, et al. Duration of device-based fever prevention after cardiac arrest. N Engl J Med. 2023;388(10):888-897.

Context & Rationale

Background
- Post-cardiac arrest brain injury evolves over hours to days; temperature is a modifiable physiological variable plausibly linked to secondary injury.
- Clinical practice shifted from routine deep hypothermia towards active temperature management aimed at avoiding fever, particularly after rewarming.
- Guidelines have recommended active fever prevention for prolonged periods (often framed as up to 72 hours), but the evidential basis for the duration component has been indirect and largely observational.
- Device-based temperature control is resource intensive and may extend sedation, ventilation, and ICU workload; the clinical value of extending active fever prevention beyond early post-ROSC care was uncertain.
Research Question/Hypothesis
- In comatose survivors of out-of-hospital cardiac arrest (presumed cardiac cause) managed with device-based temperature control at 36°C for 24 hours, does extending device-based fever prevention to 72 hours (vs 36 hours) improve survival and neurological outcomes?
- Hypothesis: 72-hour device-based fever prevention would reduce the composite of death or poor neurological outcome compared with 36-hour management.
Why This Matters
- Defines whether “duration” of active fever prevention is a causal component of post-arrest temperature management or simply a practice convention.
- Informs ICU pathway standardisation: prolonged device use, sedation exposure, and nursing workload must be justified by patient-centred benefit.
- Clarifies whether post-arrest fever is more likely a mediator (treatable) versus a marker of injury severity (less modifiable) when fever prevention is already routine.

Design & Methods

Research Question: Among comatose adult OHCA survivors treated with device-based temperature control at 36°C for 24 hours, does continuing device-based fever prevention for 72 hours (vs 36 hours) reduce death or poor neurological outcome?
Study Type: Investigator-initiated, randomised, controlled, open-label, multicentre (two Danish centres), embedded as a second-stage randomisation within the BOX platform trial; web-based randomisation with variable block sizes; stratified by site.
Population:
- Setting: ICU-based post-cardiac arrest care at Rigshospitalet (Copenhagen University Hospital) and Odense University Hospital (Denmark).
- Key inclusion: Adults ≥18 years; out-of-hospital cardiac arrest; sustained ROSC; presumed cardiac cause; unconscious after ROSC (GCS ≤8); eligible for targeted temperature management and BOX trial inclusion.
- Key exclusions: In-hospital cardiac arrest; non-cardiac cause; pregnancy; unwitnessed asystole; suspected intracranial bleeding or stroke; additional protocol-specified exclusions (details not fully reported in the index manuscript).
Intervention:
- Common initial phase (both groups): Device-based temperature management to 36°C for 24 hours.
- Rewarming: To 37°C at a maximum rate of 0.5°C/hour.
- 72-hour group (longer duration): Continue device-based fever prevention (target 37°C) for 48 hours after rewarming (total temperature management duration 72 hours from initiation), or until awakening.
- Post-device fever treatment: If temperature >38°C after discontinuation, treat with paracetamol and physical measures (uncovering).
Comparison:
- Common initial phase (both groups): Device-based temperature management to 36°C for 24 hours.
- Rewarming: To 37°C at a maximum rate of 0.5°C/hour.
- 36-hour group (shorter duration): Continue device-based fever prevention (target 37°C) for 12 hours after rewarming (total temperature management duration 36 hours from initiation), or until awakening.
- Post-device fever treatment: If temperature >38°C after discontinuation, treat with paracetamol and physical measures (uncovering).
Blinding: Open-label (no blinding of clinicians or participants); primary and secondary outcomes derived from discharge status and follow-up assessments, with potential for detection bias in functional outcomes.
Statistics: No dedicated prospective sample size calculation for this substudy; with 789 participants, protocol anticipated 80% power at two-sided alpha 0.05 to detect a 27.5% relative difference in the primary composite outcome given the observed BOX event rate; primary analysis was intention-to-treat using Cox proportional hazards models (hazard ratios with 95% CI).
Follow-Up Period: Primary outcome assessed at discharge within 90 days; neurological and cognitive follow-up at approximately 3 months where feasible.

Key Results

This trial was not stopped early. Follow-up and analyses proceeded per the prespecified BOX platform structure.

Outcome	36-hour fever prevention	72-hour fever prevention	Effect	p value / 95% CI	Notes
Primary: death or CPC 3–4 at discharge (within 90 days)	127/393 (32.3%)	133/396 (33.6%)	HR 0.99	95% CI 0.77 to 1.26; P=0.92	HR reported in manuscript (direction not explicitly restated; table reports HR with 36h and 72h columns).
Death within 90 days	116/393 (29.5%)	120/396 (30.3%)	HR 0.97	95% CI 0.75 to 1.26; P=0.83	All-cause mortality.
Modified Rankin scale at 3 months (median, IQR)	1 (0–2); assessed n=255	1 (0–2); assessed n=251	Not reported	Not reported	Functional outcome among those assessed; follow-up completeness limited by death and logistics.
Montreal Cognitive Assessment at 3 months (median, IQR)	26 (22–28); assessed n=256	27 (24–29); assessed n=253	Not reported	Not reported	Cognitive outcome among those assessed.
Temperature >37.7°C at ≥1 time point (24–72 hours)	197/393 (50.1%)	151/396 (38.1%)	RR 1.28	95% CI 1.10 to 1.49; P not reported	RR reflects higher fever exposure in the 36-hour group (numerator = 36h).
Temperature >38.5°C at ≥1 time point (24–72 hours)	46/393 (11.6%)	24/396 (6.1%)	RR 1.49	95% CI 1.06 to 2.08; P not reported	RR reflects higher fever exposure in the 36-hour group (numerator = 36h).
Any serious adverse event	170/393 (43.3%)	179/396 (45.2%)	Not reported	Not reported	Composite SAE reporting; individual categories below.
Infection (serious adverse event)	152/393 (38.7%)	162/396 (40.9%)	Not reported	Not reported	No signal of higher infection rate with longer device-based fever prevention.
Cardiac arrhythmia requiring treatment (serious adverse event)	123/393 (31.3%)	117/396 (29.5%)	Not reported	Not reported	High baseline arrhythmia burden post-OHCA; similar between groups.
Bleeding requiring transfusion (serious adverse event)	50/393 (12.7%)	51/396 (12.9%)	Not reported	Not reported	No difference observed.
Seizure requiring treatment (serious adverse event)	22/393 (5.6%)	15/396 (3.8%)	Not reported	Not reported	Event counts small; interpretation limited.

Extending device-based fever prevention to 72 hours improved temperature separation (e.g., temperature >37.7°C: 50.1% in 36-hour group vs 38.1% in 72-hour group; RR 1.28; 95% CI 1.10 to 1.49), but did not translate into improved clinical outcomes.
The primary composite outcome was neutral (32.3% vs 33.6%; HR 0.99; 95% CI 0.77 to 1.26; P=0.92), and mortality at 90 days was similar (29.5% vs 30.3%; HR 0.97; 95% CI 0.75 to 1.26; P=0.83).
No clear safety signal emerged: any serious adverse event occurred in 43.3% vs 45.2% (36-hour vs 72-hour).

Internal Validity

Randomisation and allocation: Web-based randomisation with variable block sizes and stratification by site; allocation concealment is likely for assignment generation, but treatment was unblinded after allocation.
Post-randomisation exclusions / non-receipt: 7/393 (1.8%) in the 36-hour group and 10/396 (2.5%) in the 72-hour group did not receive the assigned intervention after randomisation, diluting separation potential.
Performance and detection bias: Open-label temperature duration may influence sedation, ventilation, mobilisation, and discharge decisions; CPC at discharge and functional outcomes are susceptible to non-blinded assessment pathways.
Protocol adherence and separation of the variable of interest:
- Temperature separation was demonstrable but incomplete: temperature >37.7°C at 24–72 hours occurred in 50.1% (36-hour) vs 38.1% (72-hour); RR 1.28; 95% CI 1.10 to 1.49.
- Higher-grade fever was less common overall but still more frequent with shorter duration: temperature >38.5°C at 24–72 hours occurred in 11.6% vs 6.1%; RR 1.49; 95% CI 1.06 to 2.08.
- After device discontinuation, fever >38°C was treated in both groups with paracetamol and physical measures, potentially attenuating between-group exposure differences.
Baseline characteristics: Groups were closely balanced and predominantly “cardiac-arrest-with-shockable-rhythm” phenotype: age 63±12 years in both; male 80.7% vs 81.1%; shockable initial rhythm 83.2% vs 82.6%; median time to ROSC 20 min (IQR 14–30 vs 14–31); emergency coronary angiography 88.3% vs 88.6%.
Heterogeneity: Two-centre design limits site-level heterogeneity; device choice differed by centre (surface vs intravascular), which could introduce performance variability, but stratified randomisation mitigates systematic imbalance.
Timing: Randomisation occurred after ROSC and early stabilisation; median time from arrest to randomisation was 146 minutes (IQR 113–186 in 36-hour group vs 112–189 in 72-hour group), implying a pragmatic but not ultra-early initiation point.
Dose: The tested “dose” was duration of device-based fever prevention (36 vs 72 hours) rather than a differential temperature target; clinically relevant benefit would require that prolonging active normothermia meaningfully changes pathophysiology beyond what antipyretics and routine care already achieve.
Outcome assessment: Primary endpoint uses CPC at discharge (within 90 days), which is clinically meaningful but coarse; discharge timing and withdrawal-of-life-sustaining-therapy pathways may influence the endpoint.
Statistical rigour: Intention-to-treat approach with hazard ratios and confidence intervals; protocol power assumption targeted a large relative effect (27.5%), so smaller effects remain statistically possible.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong: randomisation and baseline balance are robust and temperature separation occurred, but open-label delivery and functional outcome ascertainment introduce meaningful risks of performance and detection bias, particularly for non-mortality outcomes.

External Validity

Population representativeness: Predominantly witnessed OHCA with high bystander CPR and shockable rhythms; results best generalised to similar “presumed cardiac cause” cohorts with organised post-arrest systems of care.
Important exclusions: Non-cardiac causes, in-hospital arrests, and selected severe neurological presentations were excluded; findings should not be extrapolated to broader arrest aetiologies or in-hospital cardiac arrest populations.
Applicability across systems: Denmark’s structured post-resuscitation pathways and high access to coronary angiography (≈88%) may not reflect resource-limited environments, potentially modifying baseline prognosis and co-intervention patterns.
Intervention transportability: Both arms used device-based temperature management and fever treatment protocols; applicability is strongest for centres already using similar devices and a 36°C/rewarm-to-37°C framework.

Conclusion on External Validity: Generalisability is good for high-resource ICUs treating comatose OHCA of presumed cardiac cause with established post-arrest pathways, but is limited for non-cardiac, in-hospital, and predominantly non-shockable cohorts.

Strengths & Limitations

Strengths:
- Clinically direct question addressing a high-variation component of standard post-arrest care (duration of fever prevention).
- Large sample (n=789) within a rigorously run platform framework, with excellent baseline balance and pragmatic delivery.
- Demonstrated biological separation in fever exposure (higher temperature excursions in the shorter-duration group).
- Patient-centred endpoints including mortality and functional/cognitive measures (mRS and MoCA at ~3 months) among those assessed.
Limitations:
- Open-label design; potential co-intervention differences (sedation duration, ventilation, mobilisation, and discharge timing) cannot be excluded.
- Primary endpoint incorporates CPC at discharge (within 90 days), which may be influenced by local discharge practices and is an imprecise neurological scale.
- Follow-up assessments (mRS, MoCA) were available only for a subset of the randomised cohort, limiting interpretability for survivorship outcomes.
- Power assumptions were geared toward detecting a large effect (27.5% relative difference); smaller clinically meaningful effects may be missed.
- Two-country, two-centre context may not reflect broader international variation in post-arrest care bundles and prognostication practices.

Interpretation & Why It Matters

Clinical practice implication

In patients already managed with device-based temperature control to 36°C for 24 hours and active fever treatment thereafter, extending device-based fever prevention from 36 to 72 hours did not improve the composite of death or poor neurological outcome, nor 90-day mortality.
Mechanistic interpretation

Despite reduced fever exposure with longer device use, the absence of outcome benefit supports the interpretation that modest differences in post-arrest fever exposure (within a framework that already treats fever >38°C) may be insufficient to alter clinically important brain injury trajectories.
Systems and resource significance

If confirmed across settings, shortening device-based fever prevention duration could reduce device-hours and nursing workload; however, effects on sedation duration, ventilation time, and ICU length of stay were not reported and remain implementation-relevant unknowns.

Controversies & Subsequent Evidence

Guideline-recommended duration vs direct evidence: The trial directly tested a key “duration” element of temperature management that had been operationalised in guidelines despite limited direct RCT evidence for a specific duration, particularly beyond early post-ROSC care.¹
What “fever prevention” actually means in practice: Both groups received active fever treatment after device discontinuation (paracetamol and physical measures), so the causal contrast is “device-based prevention duration” rather than “fever treated vs not treated”; this may partly explain why substantial temperature separation did not yield clinical benefit.
Endpoint construction and susceptibility to care-pathway effects: CPC at discharge is clinically meaningful but coarse and can be influenced by discharge timing and withdrawal-of-life-sustaining-therapy practices; neutrality could reflect true absence of effect or modest effect masked by endpoint and pathway variability.
Platform-trial context and bundle-of-care complexity: The BOX programme tested multiple physiological targets; a key methodological implication is that modest incremental differences in a single physiological domain may yield small (or absent) net effects when embedded in high-quality post-arrest care, reinforcing the need for precise, adequately powered tests of each component.²
Subsequent evidence direction: The neutral clinical results in the presence of improved fever suppression are consistent with a broader trend in post-arrest temperature research: meticulous avoidance of fever is prioritised, while incremental intensification (deeper targets or longer durations) has not consistently produced outcome gains in contemporary care bundles.

Summary

Randomised, open-label BOX substudy (n=789) comparing 36-hour vs 72-hour device-based fever prevention after OHCA in comatose adults, after a shared 24-hour 36°C phase and standardised rewarming.
Primary outcome (death or CPC 3–4 at discharge within 90 days) was neutral: 32.3% (36-hour) vs 33.6% (72-hour); HR 0.99; 95% CI 0.77 to 1.26; P=0.92.
90-day mortality was similar: 29.5% vs 30.3%; HR 0.97; 95% CI 0.75 to 1.26; P=0.83.
Longer device use reduced fever exposure between 24–72 hours (e.g., temperature >37.7°C: 50.1% vs 38.1%; RR 1.28; 95% CI 1.10 to 1.49), but without detectable patient-centred benefit.
Serious adverse events were common but similar between groups (any SAE: 43.3% vs 45.2%), with no clear safety trade-off from longer fever prevention.

Overall Takeaway

In comatose survivors of presumed cardiac-cause OHCA already receiving device-based temperature control at 36°C for 24 hours and active fever treatment thereafter, extending device-based fever prevention to 72 hours reduced fever exposure but did not improve survival or neurological outcomes. The trial reframes “72-hour fever prevention” as a practice preference rather than an evidence-based determinant of outcome, at least within contemporary post-arrest care bundles and with fever treatment available after device discontinuation.

Overall Summary

Longer device-based fever prevention improved temperature separation but did not change death or neurological disability at discharge within 90 days.
For many ICUs, a 36-hour strategy with consistent post-device fever treatment appears a reasonable default, pending local implementation considerations (sedation/workload) and guideline evolution.

Foundational Trials