← Back
NeuroTrials.ai
Neurology Clinical Trial Database

INTERACT3

The third Intensive Care Bundle with Blood Pressure Reduction in Acute Cerebral Haemorrhage Trial (INTERACT3): an international, stepped wedge cluster randomised controlled trial

Year of Publication: 2023

Authors: Ma L, Hu X, Song L, et al.

Journal: The Lancet

Citation: Ma L, et al. Lancet. 2023;402:27-40.

Link: https://www.thelancet.com/journals/lance...0806-1/fulltext


Clinical Question

Does a goal-directed care bundle improve functional outcomes in patients with acute intracerebral hemorrhage (ICH)?


Study Overview

Objective

Determine whether a care bundle targeting multiple physiological variables improves outcomes in patients with acute intracerebral hemorrhage.

Study Summary

In patients with acute spontaneous ICH presenting within 6 hours, a goal-directed care bundle (targeting SBP <140 mm Hg, glucose, temperature, and INR control) improved 6-month functional outcomes versus usual care. This is the first phase 3 RCT to show improved outcomes from acute treatment of ICH.

Intervention

  • Stepped-wedge cluster randomized trial across 121 hospitals in 10 countries. Hospitals transitioned from usual care to a care bundle consisting of:
  • SBP <140 mm Hg within 1 hour
  • Blood glucose control (6.1–7.8 mmol/L non-DM; 7.8–10.0 mmol/L in DM)
  • Antipyrexia (temperature <37.5°C)
  • INR <1.5 (if on warfarin) Primary outcome: mRS at 6 months.

Bottom Line

A goal-directed care bundle targeting early BP lowering (<140 mmHg), glucose and temperature control, and reversal of anticoagulation improved 6-month functional outcomes in acute ICH, with a favorable shift in mRS distribution (common OR 0.86, 95% CI 0.76–0.97, p=0.015).

Major Points

  • Largest pragmatic ICH management trial: 7,036 patients randomized (6,255 in primary analysis) across 121 hospitals in 10 countries (9 LMICs plus Chile) using a stepped-wedge cluster randomized design.
  • Care bundle targeted four modifiable physiological parameters: BP <140 mmHg within 1 hour, temperature <37.5°C, glucose 6.1–7.8 mmol/L in patients without diabetes (7.8–10.0 mmol/L in those with diabetes), and rapid anticoagulant reversal (INR <1.5 within 1 hour if on warfarin).
  • Significantly improved functional outcomes: favorable shift in mRS distribution at 6 months (common OR 0.86, 95% CI 0.76–0.97, p=0.015; adjusted OR 0.84, 0.73–0.97, p=0.017).
  • Lower odds of death at 6 months in the primary model (OR 0.77, 95% CI 0.63–0.95, p=0.015; crude 13.6% vs 16.6%), though this became non-significant after adjustment for country and patient characteristics (0.84, 0.65–1.07, p=0.16).
  • Significantly fewer serious adverse events in the care bundle group (516/3221 [16.0%] vs 767/3815 [20.1%]; p=0.0098).
  • Health-related quality of life EQ-5D utility score was significantly higher in the care bundle group (mean difference 0.04, 95% CI 0.02–0.07; p=0.0008).
  • Builds on INTERACT2 (2013, intensive BP lowering in ICH) by adding glucose, temperature, and anticoagulant reversal — a comprehensive 'bundle' approach rather than single-intervention.
  • Implemented via site-based quality improvement with centralized training — a scalable model applicable to diverse healthcare systems.
  • Stepped-wedge design allowed every site to eventually receive the intervention, improving ethical acceptability.
  • Only ~1% of patients were on anticoagulants (Table 1: 0.9% care bundle vs 1.0% usual care) — the anticoagulant reversal component contributed minimally; main drivers were likely BP and glucose management.
  • Together with INTERACT2 and ATACH-2, defines the current approach to acute ICH management — INTERACT3 extended the evidence to bundled care in LMIC settings.

Design

Study Type: Pragmatic, international, multicentre, blinded endpoint, stepped-wedge cluster randomised controlled trial

Randomization: 1

Blinding: Open-label with blinded outcome assessment (PROBE)

Enrollment Period: December 2017 – December 2021 (patient recruitment)

Follow-up Duration: 6 months

Centers: 121

Countries: Brazil, China, India, Mexico, Nigeria, Pakistan, Peru, Sri Lanka, Vietnam, Chile

Sample Size: 7036

Analysis: Modified intention-to-treat with mixed-effects ordinal logistic regression (random effect for cluster, fixed effects for time period and group assignment)


Inclusion Criteria

  • Adult patients (≥18 years)
  • Presenting within 6 hours of symptom onset
  • Imaging-confirmed spontaneous intracerebral hemorrhage

Exclusion Criteria

  • ICH secondary to structural brain lesion (AVM, aneurysm, tumor, trauma, previous cerebral infarction)
  • ICH related to reperfusion therapy (thrombolysis or thrombectomy)
  • Unlikely to adhere to the study treatment or follow-up regimen per treating clinician

Arms

FieldControlCare Bundle
InterventionStandard management per local guidelines without protocolized care bundleSBP <140 mmHg within 1 h; temperature <37.5°C within 1 h; glucose 6.1–7.8 mmol/L (non-diabetic) or 7.8–10.0 mmol/L (diabetic); INR <1.5 within 1 h if on warfarin (fresh frozen plasma or prothrombin complex concentrate)
Duration6 months follow-up6 months follow-up

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Functional recovery measured by modified Rankin Scale (mRS) at 6 months, analysed as an ordinal shift across all categories (common odds ratio from ordinal logistic regression)PrimaryMore unfavorable distribution of mRSShift toward lower (better) mRS scorescommon OR 0.86 (unadjusted); common OR 0.84 (adjusted)0.015 (unadjusted); 0.017 (adjusted)
Death at 6 monthsSecondary16.6% (crude); 14.3% (adjusted)13.6% (crude); 11.4% (adjusted)OR 0.77 (0.63–0.95) primary model; 0.84 (0.65–1.07) fully adjusted0.015 (primary); 0.16 (fully adjusted)
Death or major disability at 6 months (mRS 3–6)Secondary57.3% (crude)53.7% (crude)OR 0.89 (0.78–1.02)0.10
Serious adverse eventsSecondary20.1% (767/3815)16.0% (516/3221)0.0098
EQ-5D overall health utility score (mean difference)Secondaryreference+0.04 (0.02–0.07)0.0008
Hospital discharge by day 7SecondaryreferenceOR 0.72 (0.53–0.98)OR 0.720.034
Any Serious Adverse EventAdverse20.1%16.0%0.0098
Recurrent ICHAdverse1.0%0.9%0.64
HypoglycemiaAdverse0.4%0.6%0.49

Criticisms

  • Open-label stepped-wedge design — sites knew when they transitioned to the care bundle, potentially introducing Hawthorne effect and performance bias (outcome assessment was masked).
  • Cluster design relies on site compliance and fidelity to all four bundle components — variable implementation across 121 sites in 10 countries limits internal validity.
  • Mortality benefit in the primary model (OR 0.77, p=0.015) lost statistical significance after adjustment for country and patient characteristics (0.84, p=0.16) — raising questions about robustness.
  • Cannot determine which bundle component(s) drove the benefit — BP, glucose, temperature, or anticoagulant reversal contributions are impossible to disentangle.
  • Predominantly Asian population (China contributed the majority of patients) — generalizability to other regions' ICH populations and healthcare systems is uncertain.
  • Only ~1% on anticoagulants — the reversal component, which has the strongest individual evidence base, contributed minimally to the observed effect.
  • COVID-19 pandemic overlap (enrolment ended Dec 2021) affected care patterns, hospital resources, follow-up completeness, and site recruitment timing.
  • The proportional-odds assumption for the primary ordinal analysis was violated (treatment effect not constant across mRS categories).
  • Extended recruitment in the intervention group (due to pandemic delays) may have introduced selection bias, partially addressed by post-hoc calendar-time adjustment.

Funding

Joint Global Health Trials scheme (UK Department of Health and Social Care, Foreign, Commonwealth & Development Office, Medical Research Council, Wellcome Trust); West China Hospital; NHMRC (Australia); Sichuan Credit Pharmaceutic and Takeda China

Based on: INTERACT3 (The Lancet, 2023)

Authors: Ma L, Hu X, Song L, et al.

Citation: Ma L, et al. Lancet. 2023;402:27-40.

Content summarized and formatted by NeuroTrials.ai.