Pi Is 0140673623008061

Articles
The third Intensive Care Bundle with Blood Pressure

Reduction in Acute Cerebral Haemorrhage Trial (INTERACT3):
an international, stepped wedge cluster randomised
controlled trial
Lu Ma*, Xin Hu*, Lili Song*, Xiaoying Chen*, Menglu Ouyang, Laurent Billot, Qiang Li, Alejandra Malavera, Xi Li, Paula Muñoz-Venturelli,
Asita de Silva, Nguyen Huy Thang, Kolawole W Wahab, Jeyaraj D Pandian, Mohammad Wasay, Octavio M Pontes-Neto, Carlos Abanto,
Antonio Arauz, Haiping Shi, Guanghai Tang, Sheng Zhu, Xiaochun She, Leibo Liu, Yuki Sakamoto, Shoujiang You, Qiao Han, Bernard Crutzen,
Emily Cheung, Yunke Li, Xia Wang, Chen Chen, Feifeng Liu, Yang Zhao, Hao Li, Yi Liu, Yan Jiang, Lei Chen, Bo Wu, Ming Liu, Jianguo Xu, Chao You,
Craig S Anderson, for the INTERACT3 Investigators†
Summary
Background Early control of elevated blood pressure is the most promising treatment for acute intracerebral Lancet 2023; 402: 27–40
haemorrhage. We aimed to establish whether implementing a goal-directed care bundle incorporating protocols for Published Online
early intensive blood pressure lowering and management algorithms for hyperglycaemia, pyrexia, and abnormal May 25, 2023
https://doi.org/10.1016/
anticoagulation, implemented in a hospital setting, could improve outcomes for patients with acute spontaneous
S0140-6736(23)00806-1
intracerebral haemorrhage.
See Comment page 2
*Joint first authors
Methods We performed a pragmatic, international, multicentre, blinded endpoint, stepped wedge cluster
†INTERACT3 Investigators listed
randomised controlled trial at hospitals in nine low-income and middle-income countries (Brazil, China, India,
in the appendix (pp 4–9)
Mexico, Nigeria, Pakistan, Peru, Sri Lanka, and Viet Nam) and one high-income country (Chile). Hospitals were
Department of Neurosurgery
eligible if they had no or inconsistent relevant, disease-specific protocols, and were willing to implement the care (Prof L Ma MD, X Hu MD,
bundle to consecutive patients (aged ≥18 years) with imaging-confirmed spontaneous intracerebral haemorrhage X Li MM, Prof H Li MD,
presenting within 6 h of the onset of symptoms, had a local champion, and could provide the required study data. Prof Y Liu MD, Prof J Xu MD,
Prof C You MD,
Hospitals were centrally randomly allocated using permuted blocks to three sequences of implementation,
Prof C S Anderson PhD) and
stratified by country and the projected number of patients to be recruited over the 12 months of the study period. Department of Neurology
These sequences had four periods that dictated the order in which the hospitals were to switch from the control (Prof L Chen PhD, Prof B Wu MD,
usual care procedure to the intervention implementation of the care bundle procedure to different clusters of Prof M Liu PhD), West China
Hospital, Sichuan University,
patients in a stepped manner. To avoid contamination, details of the intervention, sequence, and allocation periods
Chengdu, China; The George
were concealed from sites until they had completed the usual care control periods. The care bundle protocol Institute for Global Health
included the early intensive lowering of systolic blood pressure (target <140 mm Hg), strict glucose control (target China, Beijing, China
6·1–7·8 mmol/L in those without diabetes and 7·8–10·0 mmol/L in those with diabetes), antipyrexia treatment (L Song PhD, M Ouyang PhD,
Y Li MD, C Chen MD, Y Zhao PhD,
(target body temperature ≤37·5°C), and rapid reversal of warfarin-related anticoagulation (target international
Prof C S Anderson); The George
normalised ratio <1·5) within 1 h of treatment, in patients where these variables were abnormal. Analyses were Institute for Global Health,
performed according to a modified intention-to-treat population with available outcome data (ie, excluding sites Faculty of Medicine, University
that withdrew during the study). The primary outcome was functional recovery, measured with the modified of New South Wales, Sydney,
NSW, Australia (L Song,
Rankin scale (mRS; range 0 [no symptoms] to 6 [death]) at 6 months by masked research staff, analysed using X Chen PhD, M Ouyang,
proportional ordinal logistic regression to assess the distribution in scores on the mRS, with adjustments for Prof L Billot MRes, Q Li MBiostat,
cluster (hospital site), group assignment of cluster per period, and time (6-month periods from Dec 12, 2017). This A Malavera MD,
trial is registered at Clinicaltrials.gov (NCT03209258) and the Chinese Clinical Trial Registry (ChiCTR-IOC-17011787) Prof P Muñoz-Venturelli PhD,
L Liu MSEng, X Wang PhD,
and is completed. C Chen, Prof C S Anderson);
Clinical Research Center,
Findings Between May 27, 2017, and July 8, 2021, 206 hospitals were assessed for eligibility, of which 144 hospitals in Faculty of Medicine Clinica
ten countries agreed to join and were randomly assigned in the trial, but 22 hospitals withdrew before starting to Alemana Universidad del
Desarrollo, Santiago, Chile
enrol patients and another hospital was withdrawn and their data on enrolled patients was deleted because regulatory (Prof P Muñoz-Venturelli,
approval was not obtained. Between Dec 12, 2017, and Dec 31, 2021, 10 857 patients were screened but 3821 were Prof C S Anderson); Clinical Trials
excluded. Overall, the modified intention-to-treat population included 7036 patients enrolled at 121 hospitals, with Unit, Faculty of Medicine,
3221 assigned to the care bundle group and 3815 to the usual care group, with primary outcome data available in University of Kelaniya,
Colombo, Sri Lanka
2892 patients in the care bundle group and 3363 patients in the usual care group. The likelihood of a poor functional (Prof A de Silva DPhil); Stroke
outcome was lower in the care bundle group (common odds ratio 0·86; 95% CI 0·76–0·97; p=0·015). The favourable Unit, 115 Hospital,
shift in mRS scores in the care bundle group was generally consistent across a range of sensitivity analyses that Ho Chi Minh City, Viet Nam
included additional adjustments for country and patient variables (0·84; 0·73–0·97; p=0·017), and with different (Prof N H Thang MD);
Department of Medicine,
approaches to the use of multiple imputations for missing data. Patients in the care bundle group had fewer serious University of Ilorin & University
adverse events than those in the usual care group (16·0% vs 20·1%; p=0·0098).
www.thelancet.com Vol 402 July 1, 2023 27

Articles
of Ilorin Teaching Hospital, Interpretation Implementation of a care bundle protocol for intensive blood pressure lowering and other management
Ilorin, Nigeria algorithms for physiological control within several hours of the onset of symptoms resulted in improved functional
(Prof K W Wahab MD);
outcome for patients with acute intracerebral haemorrhage. Hospitals should incorporate this approach into clinical
Neurology Department,
Christian Medical College and practice as part of active management for this serious condition.
Hospital, Ludhiana, India
(Prof J D Pandian MD); Funding Joint Global Health Trials scheme from the Department of Health and Social Care, the Foreign,
Department of Medicine, The
Commonwealth & Development Office, and the Medical Research Council and Wellcome Trust; West China Hospital;
Aga Khan University, Karachi,
Pakistan (Prof M Wasay MD); the National Health and Medical Research Council of Australia; Sichuan Credit Pharmaceutic and Takeda China.
Department of Neurology,
Ribeirão Preto Medical School, Copyright © 2023 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0
University of São Paulo,
São Paulo, Brazil
license.
(Prof O M Pontes-Neto PhD);
The Cerebrovascular Disease Introduction results across randomised controlled trials8,9 and little
Research Center, National Intracerebral haemorrhage is the most serious and least evidence specifically in patients with a large intracerebral
Institute of Neurological
Sciences, Lima, Peru
treatable form of stroke and accounts for approxi haemorrhage or who require neurosurgical intervention10
(Prof C Abanto MD); Instituto mately 20% of the almost 20 million new strokes that has restricted the uptake of this strategy in clinical
Nacional de Neurologia y occur globally each year.1 Most cases of intracerebral practice where guidelines generally provide an inter
Neurocirugia Manuel Velasco haemorrhage occur in low-income and middle-income mediate strength to their recommendation.3–7 Moreover,
Suarez, Mexico City, Mexico
(Prof A Arauz PhD); Department
countries (LMICs), where there is a high prevalence of the absence of a proven medical or surgical treatment for
of Neurosurgery, Suining hypertension, unhealthy diets (eg, a high salt intake), and intracerebral haemorrhage has resulted in an absence of
Central Hospital, Suining, other risk factors.2 Because elevated blood pressure is urgency to treat these patients and a low threshold for the
China (H Shi MM); Department common after the onset of intracerebral haemorrhage withdrawal of active care in these patients,11,12 which
of Neurology, Liaoning
Thrombus Treatment Centre of
and is strongly associated with a poor outcome, a central contrasts sharply with modern systems of care for
Integrated Chinese and component of the management of patients is to provide patients with acute ischaemic stroke.
Western Medicine, Shenyang, treatment to lower blood pressure towards a systolic Efforts to improve the success of randomised controlled
China (G Tang MBBS); target of 140 mm Hg or less.3–7 However, inconsistent trials in identifying an effective treatment for intracerebral
Department of Neurosurgery,
Dazhu County People’s
Hospital, Dazhou, China
(S Zhu MBBS); Department of Research in context
Neurosurgery, Jiangsu Rudong
County People’s Hospital,
Evidence before this study implementation of an intensive care bundle in acute
Nantong, China (X She MM); We searched PubMed (from Jan 1, 1970, to Oct 30, 2022) and intracerebral haemorrhage.
Department of Neurology, Embase (from Jan 1, 1947, to Oct 30, 2022), with no language
Graduate School of Medicine, Added value of this study
or data restrictions, on Nov 24, 2022, for publications with
Nippon Medical School, Tokyo, The third Intensive Care Bundle with Blood Pressure
Japan (Y Sakamoto MD); relevant text words in the title or abstract or keywords that
Reduction in Acute Cerebral Haemorrhage Trial is the only
Department of Neurology and included: “intracerebral haemorrhage” OR “haemorrhagic
randomised controlled trial involving a care bundle that
Clinical Research Center of stroke”, “care bundle” OR “treatment combination”, “blood
Neurological Disease, The included intensive blood pressure lowering treatment in
pressure” OR “blood pressure lowering”, “blood glucose” OR
Second Affiliated Hospital of acute intracerebral haemorrhage. The primary result was that
Soochow University, Suzhou, “hyperglycaemia”, “body temperature”, and “anticoagulation
the implementation of the care bundle across participating
China (S You MD); Department reversal”. Studies were eligible for inclusion if they assessed
hospitals resulted in patients having a better functional
of Neurology, Suzhou Hospital the effectiveness of a combination of treatments on clinical
of Traditional Chinese outcome (measured on the modified Rankin Scale) at
outcomes. We identified only one completed cluster
Medicine, Suzhou, China 6 months post-treatment compared with usual care.
(Q Han MD); Department of randomised controlled trial, done in 2011, the Quality in Acute
This result included favourable effects on survival and health-
Radiology, Cliniques Stroke Care trial, which showed that a combined protocol to
related quality of life. The results could not be explained by
Universitaires Saint-Luc, manage fever, hyperglycaemia, and swallowing dysfunction
Brussels, Belgium temporal trends in the characteristics of the patients or their
improved the functional outcome in patients with acute
(B Crutzen MD); Department of management.
Radiology, Grand Hôpital de stroke (both intracerebral haemorrhage and ischaemic stroke).
Charleroi, Charleroi, Belgium A 2019 study in Salford, UK, showed a significant association Implications of all the available evidence
(B Crutzen); Neurology between the implementation of a care bundle of A simple, time-sensitive, goal-directed, care bundle protocol,
Department, Royal Prince
anticoagulation reversal, blood pressure control, and rapid with a foundation strategy of early intensive blood pressure
Alfred Hospital, Sydney,
Australia (E Cheung MD, access to neurosurgery and improved survival after cerebral management to a systolic blood pressure target of less than
Prof C S Anderson); Department haemorrhage. The 2011 trial did not include blood pressure 140 mm Hg, was safe and effective in improving the
of Neurology, Shanghai East lowering in the protocol and there were few cases of functional outcome of acute intracerebral haemorrhage.
Hospital, Tongji University,
Shanghai, China (C Chen,
intracerebral haemorrhage, whereas the 2019 study was These results provide a clear implication for the rapid control
F Liu MM); Department of limited by its before-and-after design. No ongoing trials were of blood pressure and other physiological variables to be
Nursing and Evidence-based identified through a search of registered trials at ClinicalTrials. incorporated into clinical practice as a part of active
Nursing Center, West China gov. There was no strong evidence to recommend the management for this serious disease.
Hospital, Sichuan University,
28 www.thelancet.com Vol 402 July 1, 2023

Articles
haemorrhage have included the use of clinical and followed the Consolidated Standards of Reporting Trials Chengdu, Sichuan, China
imaging variables to define a potential responder group Extension reporting guideline for stepped wedge cluster (Prof Y Jiang PhD); Heart Health
Research Center, Beijing, China
with a high likelihood of early neurological deterioration randomised trials.20 Details of the protocol and statistical (Prof C S Anderson)
from ongoing haemorrhage or haematoma growth.13–15 analysis plan have been published elsewhere21,22 and are
Correspondence to:
Another effort has been to extend the assessment of available in the appendix (pp 89–221). Prof Craig S Anderson,
functional outcome beyond the conventional 90 days, The study was approved by the ethics committees at The George Institute for Global
because recovery from intracerebral haemorrhage takes participating hospitals and appropriate regulatory author Health, Faculty of Medicine,
University of New South Wales,
longer than from acute ischaemic stroke.16 Assessing ities. A mixed consent process was applied, whereby a
Sydney, NSW 2050, Australia
combinations of interventions as part of a multifaceted cluster guardian (an appropriate senior delegate [eg, chief canderson@georgeinstitute.
care bundle might also offer advantages, as shown in an executive officer]) provided consent for patients with acute org.au
Australian cluster clinical trial where the implementation intracerebral haemorrhage to receive the intervention as or
of a treatment protocol for hyperglycaemia, fever, and part of routine care. Participants (or an approved Prof Chao You, Department of
dysphagia improved outcome from acute stroke.17 Further surrogate) provided written informed consent for the Neurosurgery, West China
support for this approach was provided by a single-site, collection of sociodemographic, medical, and clinical Hospital, Sichuan University,
Chengdu 610041, China
so-called before-and-after study in the UK, where the information, and for them to be contacted to assess their youchao@vip.126.com
implementation of a quality improvement protocol that outcome at 6 months. Two amendments were made to the See Online for appendix
included the reversal of anticoagulation, intensive blood original protocol dated July 12, 2017, to document
pressure lowering, and rapid triage to neurosurgery and additional sources of funding, extend the time for each
critical care was associated with a lower 30-day case intervention period to 4 months for sites in Asia to
fatality after intracerebral haemorrhage.18 A post-hoc maximise the recruitment of patients, allow a transition
analysis of the second phase of the international Intensive time of 7–10 days before the sites entered the
Blood Pressure Reduction in Acute Cerebral Haemorrhage implementation period, for stratification variables to be
Trial (INTERACT2)19 showed that higher scores assigned incorporated into the randomisation programme, and to
for any elevation in the baseline of systolic blood pressure, extend the study timelines (April 8, 2018); and additionally
glucose, body temperature, and previous use of to expand the number of hospitals, further extend the
anticoagulants in participants independently predicted a study timelines, embed a process evaluation and economic
poor functional outcome after intracerebral haemorrhage. evaluation within the trial, clarify the inclusion criteria for
These issues informed the design of the third Intensive the diagnosis of intracerebral haemorrhage, include a
Care Bundle with Blood Pressure Reduction in Acute form to document consent in patients who withdrew from
Cerebral Haemorrhage Trial (INTERACT3), with the aim the study, and clarify the patient recruitment targets
of establishing whether a goal-directed care bundle within the study periods for the required sample size
protocol, comprising early intensive lowering of blood (Aug 12, 2019).
pressure with other management protocols for abnormal
physiological variables, improves functional outcome in a Clusters and participants
broad range of patients with acute spontaneous intra Hospital sites were approached through neurosurgery
cerebral haemorrhage. and neurology professional networks. To be eligible, they
either had no or inconsistent protocols for managing
Methods abnormal physiological variables in patients with acute
Study design intracerebral haemorrhage and had to be willing to
INTERACT3 was an international, multicentre, implement the required interventional care bundle
prospective, stepped wedge, cluster randomised, blinded, protocol as part of routine care. Eligible sites needed to
outcome assessed, controlled trial undertaken at enrol consecutive adult (age ≥18 years) patients presenting
hospitals located in nine LMICs (Brazil, China, India, within 6 h after the onset of acute intracerebral
Mexico, Nigeria, Pakistan, Peru, Sri Lanka, and haemorrhage, have a local representative (known as a
Viet Nam), and one high-income country (Chile). The champion) who was willing to lead the implementation of
trial had a hybrid discovery–implementation design, in the intervention, and be able to provide the required
which clusters of patients were prospectively followed up study data. Key patient exclusion criteria included the
to establish their outcome as their hospitals were intracerebral haemorrhage being secondary to a structural
randomly allocated to switch from control usual care to abnormality in the brain (eg, an arteriovenous
the intervention implementation at different timepoints. malformation, intracranial aneurysm, tumour, trauma,
The experimental approach was done to mirror the or previous cerebral infarction) or to reperfusion
natural process of rolling out a new quality improvement therapy (eg, intravenous thrombolysis or endovascular
policy, that of implementing a goal-directed intensive thrombectomy), or that a patient was unlikely to adhere to
care bundle protocol involving the rapid correction of any the study treatment or follow-up regimen, or both, as
abnormal physiological variables (hypertension, hyper judged by the treating clinician. Full details of the
glycaemia, and pyrexia) and altered coagulation profile in inclusion and exclusion criteria are provided in the
patients with acute intracerebral haemorrhage. The trial appendix (p 14).

Articles
Randomisation and masking threshold for the cessation of treatment; intensive control
Eligible hospital sites were randomly assigned into three of elevated blood glucose with the goal of achieving a
sequences (with four periods) by use of a computer- glucose target of 6·1–7·8 mmol/L for patients without
generated list by the trial statistician (QL) using permuted diabetes and 7·8–10·0 mmol/L for patients with diabetes
blocks, stratified by country and the projected number of as soon as possible after the initiation of treatment, as
patients to be recruited per site during the study period recommended in guidelines for minimal risk of harm
(<80, 80–160, and >160 patients). In period 1, all sites from hypoglycaemia;23 treatment of pyrexia with the goal
kept to their usual care and monitoring procedures for of achieving a body temperature of less than 37·5°C within
patients with intracerebral haemorrhage. Sites assigned 1 h of initiation; and the reversal of abnormal
to sequence 1 commenced the care bundle in period 2, anticoagulation in those taking warfarin using fresh
those in sequence 2 received the intervention at period 3, frozen plasma or prothrombin concentrate complex with
and those in sequence 3 received the intervention at the goal of reaching an international normalised ratio of
period 4, as outlined in the appendix (pp 26–30, 78). The less than 1·5 within 1 h of treatment. All target
criteria that triggered sites to progress to the next period concentrations within the care bundle were to be
were either achieving the predetermined patient maintained in patients for 7 days (or until discharge or
enrolment target or reaching a time of 3 months death, should these events occur earlier). Site-specific data
(4 months in Asia where there was greater capacity for uploaded to the study database on blood pressure, glucose,
patient recruitment) from the start date for each period. temperature, and international normalised ratio served as
To avoid contamination, details of the intervention, a method to check adherence to the protocol, and also
sequence, and allocation periods were concealed from allowed feedback and further training to be offered to sites
sites until they had completed the usual care control by trial staff. All sites received monthly performance
periods. A period of 7–10 days was used to allow sites reports and were required to have staff attend at least
time for their staff to receive training before transitioning two online quality improvement meetings held repeatedly
from usual care to the intervention period. Trained during the study. A process evaluation, whereby formative
research staff masked to group allocation and patient stakeholder engagement interviews were conducted
details located in country-based central offices (in China, during the trial, allowed insight to be gained into the
Chile, Brazil, and Nigeria) or at hospitals (in the other barriers and facilitators to change systems of care and to
countries), undertook telephone assessments of the implement the protocol.24 The information also served as a
outcome of patients at 6 months post-randomisation. further guide to the training of site staff in their integration
For logistical reasons and according to available of the care bundle into practice. Further details of the care
funding, sites were activated in batches beginning bundle protocol are provided in the appendix (pp 88–170).
in China in 2017, and subsequently in the other countries To ensure continuous recruitment, each participating
in 2019 (appendix pp 26–32, 79–80). However, because of site was required to register all patients with a diagnosis of
the COVID-19 pandemic, patient recruitment was intracerebral haemorrhage into screening and enrolment
suspended in Chile, Pakistan, and Peru between March 1 logs. For enrolled patients, demographic and clinical data,
and Aug 31, 2020, and no new hospitals outside of China including the amount of neurological impairment
commenced the recruitment of patients until Nov 2, 2020. measured using the National Institutes of Health Stroke
Because the pandemic affected patient recruitment, Scale (NIHSS; range 0–42, with higher scores indicating
mainly at sites in the intervention phase, and caused an greater severity) were collected at the time of admission to
imbalance in patient numbers between the randomly hospital. Follow-up data were collected on clinical status,
assigned groups, the duration of this phase was extended management, and outcomes at days 1 and 7 of treatment
to allow as many sites as possible to participate and enrol (or discharge or death if earlier) by hospital staff, and of
patients into the intervention group. outcomes to be ascertained at 6 months by independent
staff. Brain imaging undertaken at presentation (the
Procedures diagnostic scan), and at 24 h and 7 days of treatment in a
Before hospitals began screening and recruiting patients, subsample of 1619 patients (first seven patients in each of
key staff from participating sites were trained in the study the usual care and intervention phases at each site) were
procedures at regional and individual hospital meetings. uploaded to a secure central imaging server for analysis by
In the 7–10 days after sites had completed the control trained physicians who were masked to treatment
usual care period, these staff received online training and allocation. Results of the centrally adjudicated brain
remote communication on the intensive care bundle imaging will be presented in future publications.
protocol to be applied as a system of care for eligible
patients. The components of the care bundle included: Outcomes
early intensive blood pressure management with the goal The primary outcome was functional recovery measured
of achieving a target systolic blood pressure of less than at 6 months according to the modified Rankin scale
140 mm Hg within 1 h of the initiation of treatment, with (mRS) and analysed as an ordinal outcome (shift across
a systolic blood pressure of 130 mm Hg being the all categories). The mRS is a standard global measure of

Articles
disability scored from 0 to 6, in which scores of 0–1 upon unmasking the data, we recognised that the trial
indicate a favourable outcome without or with symptoms did not conform to a standard stepped wedge design in
but no disability; scores of 2–5 indicate increasing which all clusters contemporaneously switch from one
amounts of disability (and dependency); and a score of period to the next,29 because the period lengths were not
6 indicates death. Secondary outcomes included the equal and the times at which the different clusters
dichotomous analysis of scores on the mRS crossed from control to intervention conditions varied
at 6 months: 3–6 (disability or death) versus 0–2, and widely (appendix pp 26–30, 78–79). To more appropriately
3–5 (major disability) versus 0–2 in survivors; death at account for the effect of time, we adjusted for calendar
6 months; death or neurological deterioration time instead of trial period by using 6-month calendar
at 7 days according to a change in scores on the NIHSS, periods that corresponded to the length of follow-up for
both as a continuous measure and categorised into seven the assessment of the clinical outcomes in individual
groups (<5, 5–9, 10–14, 15–19, 20–24, ≥25, and death);25 patients from the time the first patient was recruited
health-related quality of life measured using the EuroQoL (Dec 12, 2017) to the time the last patient was recruited
Group 5-Dimension Self-Report Questionnaire (Dec 31, 2021; appendix p 32).30,31
(EQ-5D-3L); and residence (own home vs other) at We undertook a series of preplanned sensitivity
6 months. Time to discharge from hospital was also analyses with various approaches to the imputation of
planned but was only recorded at day 7 to coincide with missing values,22 variable time effects across sites,32 and
the other assessments undertaken at this time, and was additional adjustment for the covariates of country
therefore censored at 7 days. The safety outcomes of all- (grouped regionally as China vs India, Pakistan,
cause and cause-specific serious adverse events were Sri Lanka, and Viet Nam vs Brazil, Peru, Chile, Mexico,
recorded according to standard definitions for the and Nigeria), pre-stroke function (estimated categorical
duration of follow-up. mRS scores of 0 to 5), age, sex, and baseline NIHSS score
to the primary model. The NIHSS score was also
Statistical analysis analysed as a continuous variable using an unadjusted
This study was designed with 90% power (α=0·05) to hierarchal linear regression model assuming a normal
detect a common odds ratio (OR) of 0·8 for worse distribution and an identity link function. We also
functional outcome at 6 months using ordinal logistic conducted post-hoc sensitivity analyses by modelling the
regression. This design required a sample size effect of time continuously using restricted cubic splines
of 8360 patients from 110 sites, on the assumption that based on 1-month and 3-month calendar time windows.
the distribution in the scores on the mRS in the usual We found the proportional-odds assumption was
care group would be similar to the control group who violated but the analyses proceeded on the understanding
received standard less-intensive blood pressure lowering that the treatment effect was not constant across
treatment in the INTERACT2 study,26 and with 5% of categories of the mRS. The result was complemented by
patients lost to follow-up. This treatment effect a graphical assessment of shifts across categories using
corresponded to a 5·6% absolute decrease in worse bar plots as well as binary analyses. As prespecified, the
functional outcome (mRS scores 3–6) from receiving the six subgroups for analysis were age, sex, country, NIHSS
care bundle, from 55·6% down to 50·0% compared with score (<15 vs ≥15), and clinician-reported volume
usual care. We assumed an intraclass correlation (<15 mL vs 15 to <30 mL vs ≥30 mL) and location (cortical
coefficient of 0·04 on the basis of the results of vs deep vs brainstem, cerebellar, or primary ventricular)
INTERACT2 and another international cluster of the haematoma reported by investigators at baseline,
randomised trial in acute stroke.25 The required sample with tests of interaction applied between the specific
size, which corresponded to an average of 19 patients subgroup and the treatment effect on the primary
being recruited per phase at each participating hospital outcome. A post-hoc estimation of the number needed to
(ie, 110 sites × 4 phases × 19 patients), was calculated using treat for a benefit from the treatment was made to help in
PASS software 2019, by tests for two ordered categorical the interpretation of the results.
variables,27 and by adjusting for the clusters corresponding Although two interim analyses were undertaken
to the stepped wedge design.28 during the study period, given the use of a conservative
All analyses were undertaken at the patient level with Haybittle–Peto stopping rule33 and the negligible amount
adjustment for clustering. The main analyses were of type 1 error rate spent, the significance threshold for
performed on a modified intention-to-treat basis by the primary outcome, including sensitivity analyses, was
including all participants who provided written consent still p<0·05. For the seven secondary clinical outcomes,
and had primary outcome data available, regardless of we controlled the family-wise error rate by applying
protocol adherence. The primary analysis was to be done a sequential Holm–Sidak correction34 to facilitate
using ordinal logistic regression with a random effect for an interpretation of the findings.35 Between-group
cluster (hospital site), a fixed categorical effect of time differences in interventional physiological variables were
(four periods), and a fixed effect indicating the group assessed using a repeated-measure linear mixed model
assignment of each cluster at each period. However, with adjustments for treatment, time, and

Articles
treatment-by-time interactions fixed effects, site as a

206 clusters (hospital sites) assessed for eligibility random effect, and with within-patient correlations
modelled using a repeated patient effect assuming a
compound-symmetry structure. We used SAS Enterprise
62 sites excluded (32 in China, 30 in all other
countries) Guide (version 8.2) and R (version 4.0.0 or newer) for
30 unable to gain approval statistical analysis. This trial is registered with
17 had few eligible patients
15 not interested ClinicalTrials.gov (NCT03209258) and the Chinese
Clinical Trial Registry (ChiCTR-IOC-17011787).
144 clusters block randomly assigned to three groups
Role of the funding source
The funders of the study had no role in study design,
22 sites dropped out before any patients were recruited
data collection, data analysis, data interpretation, or
writing of the report.
Usual care
122 clusters analysed
Care bundle
Results
Implementation of care bundle Between May 27, 2017, and July 8, 2021, 206 hospitals
were assessed for eligibility, of which 144 hospitals in ten
Sequence 1 Sequence 2 Sequence 3
43 clusters allocated 37 clusters allocated 42 clusters allocated countries agreed to join and were randomly assigned in
the trial, but 22 hospitals subsequently withdrew before
Period 1
any patients were enrolled. Of the remaining
41 clusters participated 36 clusters participated 39 clusters participated 122 hospitals, 43 hospitals were randomly assigned into
Average size n=16 Average size n=21 Average size n=26
Variance of sizes n=13 Variance of sizes n=18 Variance of sizes n=25 sequence 1, 37 hospitals into sequence 2, and 42 hospitals
2 did not participate 1 did not participate 3 did not participate into sequence 3. For logistical reasons, sites were
Period 2 randomly assigned in four batches. The first three
batches (30 sites in each) were for sites in China, with
33 clusters participated 34 clusters participated 33 clusters participated
Average size n=19 Average size n=15 Average size n=18 each lasting 3–4 months; batch 1 from Nov 17, 2017, to
Variance of sizes n=18 Variance of sizes n=13 Variance of sizes n=16 Sept 16, 2019; batch 2 from April 20, 2018, to June 25, 2019;
10 did not participate 3 did not participate 9 did not participate
and batch 3 from Aug 29, 2018, to Nov 2, 2019. Random
Period 3
assignment of the fourth batch, which included six sites
27 clusters participated 26 clusters participated 24 clusters participated in China from April 16 to Nov 25, 2020, and all sites
Average size n=17 Average size n=18 Average size n=18
Variance of sizes n=19 Variance of sizes n=17 Variance of sizes n=22
outside of China, occurred between March 20, 2019 and
16 did not participate 11 did not participate 18 did not participate March 31, 2021. One hospital allocated to sequence 3 was
Period 4
later withdrawn and their data deleted after the
implementation of the intervention because a necessary
26 clusters participated 25 clusters participated 29 clusters participated
Average size n=22 Average size n=21 Average size n=24 regulatory approval was not obtained. The continued
Variance of sizes n=23 Variance of sizes n=14 Variance of sizes n=23 participation of the hospitals in the trial tended to decline
17 did not participate 12 did not participate 13 did not participate
across the phases because of the effect of COVID-19
(32 sites), unavailability of eligible patients (17 sites), loss
1 hospital from sequence 3 with 28 patients excluded
of interest after the champion left the hospital (11 sites),
regulatory changes (two sites), and early achievement of
the recruitment target (two sites; appendix pp 26–30).
121 clusters analysed, 10 857 patients screened
Even so, the number of patients enrolled was balanced
across the randomised sequences and periods (figure 1;
3821 patients excluded appendix p 31). Between Dec 12, 2017, and Dec 31, 2021,
10 857 patients with acute intracerebral haemorrhage
were screened for eligibility, with 3821 considered
7036 patients were randomly assigned
ineligible most often due to having a presentation longer
than 6 h after the onset of symptoms. 7036 patients were
randomly assigned in the trial, of whom 3815 were
3221 allocated to care bundle 3815 allocated to usual care allocated to the usual care group and 3221 were allocated
to the care bundle intervention group. The stepped
222 lost to follow-up 374 lost to follow-up wedge study timeline, number of patients recruited
107 missing primary outcome 78 missing primary outcome across sequences and periods, and reasons for excluding
patients are outlined in the appendix (p 33). 374 (9·8%)
2892 included in the primary analysis 3363 included in the primary analysis patients in the usual care group and 222 (6·9%) patients
in the care bundle group were lost to follow-up, and
Figure 1: Flow diagram of participating hospitals and patients another 78 (2·0%) patients in the usual care group and

Articles
107 (3·3%) patients in the care bundle group had a

missing functional status on the mRS obtained at the Care bundle (n=3221) Usual care (n=3815)
final visit. Thus, 3363 patients in the usual care group
Age, years 61·8 (12·6) 62·1 (12·6)
and 2892 patients in the care bundle group had data
Sex
available on the primary outcome and were included in
Female 1146 (35·6%) 1387 (36·4%)
the primary efficacy analysis (figure 1). The proportions
Male 2075 (64·4%) 2428 (63·6%)
of participants with primary outcome data and the
Ethnicity
method of assessment were well balanced by group and
Han Chinese 2807 (87·1%) 3490/3814 (91·5%)
period (appendix pp 31, 34–35).
Non-Han Chinese 19 (0·6%) 34/3814 (0·9%)
Baseline demographic, medical history, and the clinical
and brain imaging characteristics of patients were well Caucasian or European 2 (0·1%) 3/3814 (0·1%)
balanced between the treatment groups (table 1; Latin American 45 (1·4%) 45/3814 (1·2%)
appendix p 36). The mean age was 62·0 (SD 12·6) years, Mixed 4 (0·1%) 2/3814 (0·1%)
2533 (36·0%) of 7036 patients were female, and Other Asian 92 (2·9%) 76/3814 (2·0%)
6350 (90·3%) were Chinese. Median scores were Indian subcontinent 194 (6·0%) 143/3814 (3·7%)
12 (IQR 9–14) on the Glasgow coma scale and 13 (IQR 7–22) Other 58 (1·8%) 21/3814 (0·6%)
on the NIHSS. The cause of the intracerebral haemorrhage Main occupation*
was presumed to be related to hypertension in Unskilled 2855/3218 (88·7%) 3289/3809 (86·3%)
6574 (94·3%) of 6972 patients and located deep in a Skilled 363/3218 (11·3%) 520/3809 (13·7%)
cerebral hemisphere in 5638 (82·3%) of 6849 patients BMI, kg/m2†
(appendix p 37). The median volume of haemorrhage was Number for which this information was 2713 3172
recorded as 15·0 mL (IQR 7·8–30·0) in 6652 patients and collected
intraventricular haemorrhage was present in 2093 (29·8%) Mean 24·1 (3·7) 24·1 (3·6)
of 7032 patients. 5160 (73·3%) of 7035 patients were Medical history
admitted directly to a neurosurgery department, and History of hypertension 2207 (68·5%) 2681/3814 (70·3%)
6320 (90·4%) of 6988 patients had a blood pressure of Previous intracerebral haemorrhage 247 (7·7%) 294/3814 (7·7%)
140 mm Hg or greater, with an overall mean systolic blood Previous ischaemic stroke 246 (7·6%) 319/3814 (8·4%)
pressure of 174·5 mm Hg (SD 28·3). 2413 (36·0%) of History of coronary artery disease 70 (2·2%) 123/3814 (3·2%)
6711 patients had elevated blood glucose and the overall History of other heart disease 142 (4·4%) 155/3814 (4·1%)
mean blood glucose concentration was 8·0 mmol/L History of atrial fibrillation 31 (1·0%) 51/3814 (1·3%)
(SD 2·9). Only 120 (1·7%) of 6930 patients had an elevated History of diabetes 338 (10·5%) 391/3814 (10·3%)
body temperature and 84 (1·2%) of 6760 patients had a History of hypercholesterolaemia 90 (2·8%) 117/3813 (3·1%)
higher international normalised ratio at presentation. Current smoker 588 (18·3%) 771/3813 (20·2%)
The proportion of patients administered any Current alcohol consumption 622 (19·3%) 772/3813 (20·2%)
intravenous blood pressure lowering drug in the first Modified Rankin scale score before onset‡
24 h was higher in the care bundle group 0 2489 (77·3%) 2894/3748 (77·2%)
(2542 of 3221 [78·9%]) than the usual care group 1 358 (11·1%) 440/3748 (11·7%)
(2703 of 3811 [70·9%]); results were similar during 2 109 (3·4%) 149/3748 (4·0%)
days 2–7 (2131 of 3188 in the care bundle group 3 89 (2·8%) 77/3748 (2·1%)
[66·8%] vs 2277 of 3775 in the usual care group [60·3%]; 4 102 (3·2%) 108/3748 (2·9%)
table 2; appendix pp 38–39). The most common 5 74 (2·3%) 80/3748 (2·1%)
intravenous agents used in the first 24 h were urapidil Medications
(3351 [61·2%]), sodium nitroprusside (1169 [21·4%]), Antihypertensive medication 1446 (44·9%) 1626/3813 (42·6%)
labetalol (663 [12·1%]), nicardipine (438 [8·0%]), and Blood glucose lowering agents 250 (7·8%) 263/3814 (6·9%)
nimodipine (432 [7·9%]) reported in 5473 patients who Statin or other lipid lowering agent 89 (2·8%) 133/3814 (3·5%)
received blood pressure lowering treatment (appendix Aspirin or other antiplatelet agent 151 (4·7%) 227/3814 (6·0%)
p 38). Mean systolic blood pressures were 148·4 mm Hg Anticoagulation agent 28 (0·9%) 37/3814 (1·0%)
(SD 21·5) at 1 h and 136·1 mm Hg (16·5) at 24 h in the Systolic blood pressure, mm Hg
care bundle group, and 154·7 mm Hg (22·5) at 1 h and
Number for which this information was 3221 3813
139·0 mm Hg (17·2) at 24 h in the usual care group collected
(adjusted mean difference at 24 h of –3·6 mm Hg; Mean 174·6 (28·2) 174·5 (28·4)
95% CI –4·5 to –2·7; p<0·0001; figure 2A). The systolic Diastolic blood pressure, mm Hg
blood pressure target of 140 mm Hg or less was achieved Number for which this information was 3221 3813
more quickly in the care bundle group than in the usual collected
care group (median, 2·3 h [IQR 0·8–8·0] vs Mean 99·7 (17·7) 99·0 (17·9)
4·0 h [1·9–16·0]; table 2). The mean diastolic blood (Table 1 continues on next page)
pressures were 86·2 mm Hg (SD 13·8) at 1 h and

Articles
(median, 1·0 h [IQR 1·0–1·0] vs 1·0 h [1·0–1·0]). Patients

Care bundle (n=3221) Usual care (n=3815)
in the care bundle group with pyrexia also had more
(Continued from previous page) frequent control of their body temperature than those in
Severity of neurological deficit by scores on the NIHSS§ the usual care group (45 of 52 [86·5%]) at a median of
Number for which this information was 3149 3693 3·0 h [IQR 1·3–9·5] vs 57 of 68 [83·8%] at a median
collected
of 4·0 h [2·0–12·3]), but there was no overall adjusted
Median (IQR) 13 (7–23) 13 (6–22)
mean difference over 24 h between the two groups
Consciousness by scores on the GCS¶ (figure 2D). There were no clear differences in the
Number for which this information was 3219 3810 reversal of anticoagulation (12 of 25 [48·0%] at a median
collected
of 27 h [IQR 24·0–48·0] in the care bundle group and
Median (IQR) 12 (9–14) 12 (9–14)
24 of 59 [40·7%] at a median of 25·5 h [IQR 24·0–48·0]
Brain imaging features||
in the usual care group). Table 3 (and appendix p 40) shows
Haematoma present on CT scan 3175 (98·6%) 3680/3814 (96·5%)
there were no major differences in other management
Volume of haematoma, mL
methods over 7 days, especially in the use of
Number for which this information was 3129 3523
collected
decompressive surgery, endotracheal intubation, and
early withdrawal of care. The distribution of the reported
Median (IQR) 15 (7–30) 15 (8–30)
volumes of the haematoma and NIHSS scores over time
Side of haematoma
are shown in the appendix (pp 80–81).
Left 1628/3175 (51·3%) 1775/3676 (48·3%)
Patients in the care bundle group had better scores on
Right 1410/3175 (44·4%) 1751/3676 (47·6%)
the mRS levels compared with those allocated to the
Midline 176/3175 (5·5%) 217/3676 (5·9%)
usual care group, with a common OR of 0·86 (95% CI
Location of the haematoma
0·76–0·97; p=0·015) for an average effect across all
Cortical 299/3175 (9·4%) 329/3674 (9·0%)
mRSs violated in the proportion odds assumption test
Deep 2631/3175 (82·9%) 3007/3674 (81·8%)
(p=0·0003; figure 3; appendix pp 41–43). The favourable
Cerebellum 157/3175 (4·9%) 212/3674 (5·8%)
shifts in mRS scores in the care bundle group were still
Brainstem 156/3175 (4·9%) 195/3674 (5·3%)
present when the prespecified methods of imputation
Intraventricular haemorrhage 863 (26·8%) 1230/3811 (32·3%) were used to replace missing data, except for an extreme
Presumed cause|| score of 6 (death; appendix pp 44–46). These shifts were
Hypertension vasculopathy 3043/3197 (95·2%) 3531/3775 (93·5%) also significant after further adjustment for country and
Cerebral amyloid angiopathy 167/3197 (5·2%) 178/3775 (4·7%) patient characteristics (table 3) and when using varying
Abnormal physiological variables time trends (appendix p 46). Furthermore, the results
Systolic blood pressure ≥140 mm Hg 2905 (90·2%) 3415/3767 (90·7%) were consistent with post-hoc analyses with time
Blood glucose >7·8 mmol/L in those without 1094/3175 (34·5%) 1319/3536 (37·3%) modelled using restricted cubic splines variables
diabetes and >10·0 mmol/L in those with
diabetes
(appendix p 52). When modelling time as originally
Body temperature >37·5°C 52/3214 (1·6%) 68/3716 (1·8%)
planned using study period instead of calendar time,
there was no clear difference between the randomised
International normalised ratio ≥1·5 25/3113 (0·8%) 59/3647 (1·6%)
groups in either unadjusted or adjusted analyses, and
Data are shown as mean (SD), n (%), median (IQR), or n/N (%). GCS=Glasgow coma scale. NIHSS=National Institutes of
with or without varying time trends (appendix pp 52–54).
Health Stroke Scale. *Skilled includes professional or executive, business, sales, and service jobs; unskilled includes
driver, farmer or labourer, home duties, and other jobs. †Derived from self-reported height and weight. ‡Scores on the Secondary outcome results were consistent with all
modified Rankin scale of functional recovery range from 0 (no symptoms) to 6 (death); a score of 2 or less indicates point estimates favouring the care bundle group
functional independence. The modified Rankin scale score before stroke onset was assessed by the treating physician including significant differences in mortality, health-
with the use of information obtained from patients (if possible) or their family members. §Scores on the NIHSS range
from 0 to 42, with higher scores indicating more severe neurological deficits. ¶Scores on the GCS range from
related quality of life, and time to discharge (table 3;
15 (normal) to 3 (deep coma). ||Reported by clinician investigators; multiple options recorded. appendix pp 46–48). Compared with the usual care
group, the care bundle group had lower odds of death
Table 1: Patient characteristics at baseline
over 6 months in the primary model (common OR 0·77;
95% CI 0·63–0·95; p=0·015; appendix p 82). The
78·9 mm Hg (11·5) at 24 h in the care bundle group, and difference became non-significant after adjustment for
88·3 mm Hg (14·2) at 1 h and 80·4 mm Hg (12·1) at 24 h country and patient characteristics (0·84; 0·65–1·07;
in the usual care group (adjusted mean difference at 24 h p=0·16). There was no significant difference in the odds
of –2·4 mm Hg; 95% CI –3·0 to –1·8; p<0·0001; figure of death or disability at 90 days (mRS scores 3–6,
2B). In the care bundle group, the proportion of patients 54% vs 57%; 0·89; 0·78–1·02; p=0·10) or death or
who achieved the blood glucose target control was higher neurological deterioration at 7 days (0·89; 0·77–1·03;
than in the usual care group (147 of 1094 [13·4%] vs p=0·12). Some differences were observed in health-
86 of 1319 [6·5%]), although there were minimal related quality of life between the randomised groups,
differences in adjusted mean glucose concentrations including an increase in overall utility score in the care
over 24 h (–0·5 mmol/L; 95% CI –0·8 to –0·2; figure 2C) bundle group compared with the usual care group (mean
and in the time to achieving the glycaemic control target difference 0·04, 95% CI 0·02–0·07; p=0·0008; table 3),

Articles
but these differences diminished after further elevated blood pressure together with management
adjustment. The proportion of patients discharged from algorithms for hyperglycaemia, pyrexia, and abnormal
hospital by day 7 was significantly less in the care bundle coagulation resulted in an improved functional outcome
group (OR 0·72, 95% CI 0·53–0·98; p=0·034) but there at 6 months. The positive result was consistent across a
was no difference in the place of residence at 6 months. range of sensitivity analyses and secondary outcomes.
After controlling the family-wise error for multiple
testing, only the EQ-5D-3L utility score was significantly
Care bundle (n=3221) Usual care (n=3815)
higher in the care bundle group of the seven secondary
outcomes (appendix p 54). Location of patients after admission to hospital
There was significant heterogeneity in the treatment Neurosurgery ward 2222 (69·0%) 2938/3814 (77·0%)
effect on the primary outcome in the subgroups based on Intensive care unit 463 (14·4%) 351/3814 (9·2%)
region and COVID-19 period (appendix pp 55–61, 83), Neurology ward 255 (7·9%) 260/3814 (6·8%)
which were related variables because all non-China sites Emergency department 149 (4·6%) 163/3814 (4·3%)
were activated after the COVID-19 pandemic. There was Other area 132 (4·1%) 103/3814 (2·7%)
no heterogeneity by COVID-19 for sites in China Treatment in the first 24 h
(appendix p 84). Between-group separations in the target Intravenous blood pressure lowering treatment 2542 (78·9%) 2703/3811 (70·9%)
systolic blood pressure being reached were greater in Systolic blood pressure target <140 mm Hg 2809/2905 (96·7%) 3223/3415 (94·4%)
patients recruited in the first grouped country region of reached*
India, Pakistan, Sri Lanka, and Viet Nam compared with Median time to reaching target (IQR) 2·3 (0·8 to 8·0) 4·0 (1·9 to 16·0)
those in China and the other grouped country region Intensive treatment of blood glucose 250 (7·8%) 263/3814 (6·9%)
(Brazil, Peru, Chile, Mexico, and Nigeria; appendix Type of agent used for glycaemic control
pp 85–87). However, no major regional differences Oral agents 205/250 (82·0%) 211/263 (80·2%)
between the randomised groups were apparent between Insulin 56/250 (22·4%) 72/263 (27·4%)
China and each of the grouped regions regarding other Blood glucose target of 6·1–7·8 mmol/L in 147/1094 (13·4%) 86/1319 (6·5%)
aspects of management including decompressive surgery patients without diabetes and 7·8–10·0 mmol/L
in patients with diabetes reached*
(care bundle vs usual care, 29·5% vs 28·8% in China,
Median time to reaching target (IQR), h 1·0 (1·0 to 1·0) 1·0 (1·0 to 1·0)
2·1% vs 2·3% in India, Pakistan, Sri Lanka, and Viet
Antipyrexia treatment 314/3219 (9·8%) 280/3809 (7·4%)
Nam, and 7·2% vs 7·6% in Brazil, Peru, Chile, Mexico,
Type of antipyrexia treatment
and Nigeria) and intensive care (10·9% vs 8·4% in China,
Paracetamol 39/314 (12·4%) 43/280 (15·4%)/
9·2% vs 6·6% in India, Pakistan, Sri Lanka, and
Metamizole 7/314 (2·2%) 12/280 (4·3%)
Viet Nam, and 9·3% vs 15·2% in Brazil, Peru, Chile,
Mexico, and Nigeria; appendix pp 62–64). Hypothermia cooling with calf packing 228/314 (72·6%) 166/280 (59·3%)
Overall, patients in the care bundle group had Intravenous infusion with cooled (4°C) saline 17/314 (5·4%) 35/280 (12·5%)
significantly fewer serious adverse events than the usual Other 44/314 (14·0%) 60/280 (21·4%)
care group (516/3221 [16·0%] vs 767/3815 [20·1%]; Temperature target (≤37·5°C) reached* 45/52 (86·5%) 57/68 (83·8%)
p=0·0098). A complete list of serious adverse events and Time to reach target, h 3·0 (1·3 to 9·5) 4·0 (2·0 to 12·3)
causes of deaths is provided in the appendix (pp 65–72), Correction of abnormal coagulation
as well as details regarding the causes of death International normalised ratio target 12/25 (48·0%) 24/59 (40·7%)
<1·5 reached*
(appendix pp 73–76). From the primary model, adjusted
Time to achieving target, h 27·0 (24·0 to 48·0) 25·5 (24·0 to 48·0)
frequencies indicated 54·1% (crude number, 1553/2892
Other management from day 2 to day 7
[53·7%]) of patients in the care bundle group and 56·9%
(1927/3363 [57·3%]) in the usual care group were Intravenous blood pressure lowering treatment 2131/3188 (66·8%) 2277/3775 (60·3%)
estimated with death or major disability, and 11·4% Oral blood pressure lowering treatment 2267/3188 (71·1%) 2547/3775 (67·5%)
(407/2999 [13·6%]) in the care bundle group and 14·3% Insulin 499/3188 (15·7%) 440/3775 (11·7%)
(571/3441 [16·6%]) in the usual care group were estimated Hypothermia treatment 669/3188 (21·0%) 751/3775 (19·9%)
with death alone, therefore the estimated number needed Prothrombin complex concentrate administered 332/3188 (10·4%) 289/3775 (7·7%)
to treat for the care bundle to prevent one patient was Fresh frozen plasma administered 61/3188 (1·9%) 104/3775 (2·8%)
35 (95% CI 15 to infinity) to prevent one patient from Vitamin K administered 174/3188 (5·5%) 163/3775 (4·3%)
death or major disability and 35 (17 to infinity) to prevent Decompressive surgery 844/3196 (26·4%) 1016/3775 (26·9%)
one patient from death. Mechanical ventilation 646/3196 (20·2%) 806/3775 (21·4%)
Intensive care admission 1105/3196 (34·6%) 1435/3775 (38·0%)
Discussion Assisted feeding 1749/3196 (54·7%) 1920/3775 (50·9%)
This pragmatic stepped wedge, cluster randomised, Decision to withdraw active care 21/3196 (0·7%) 26/3775 (0·7%)
controlled trial conducted in various health-care settings Data are shown as mean (SD), n (%), median (IQR), or n/N (%). *Reached care bundle targets during the 7 days of
has shown that, in patients presenting within 6 h of the treatment.
onset of acute intracerebral haemorrhage, the use of a
Table 2: Management of patients 7 days after treatment
care bundle protocol incorporating the early control of

Articles
A
180 Mean difference (95% CI): –7·3 (–8·2 to –6·5) at 1 h
175
Systolic blood pressure (mm Hg)
–3·6 ( –4·5 to –2·7) at 24 h
170
165
160
155
150
145
140 Control
135 Intervention
130
R 15 20 45 1 2 3 4 8 12 16 20 24 2 3 4 5 6 7
Minutes Hours Days
Number at risk
Control 3415 2261 2419 2204 2853 2920 2809 3038 3134 3112 3084 3054 3028 3322 3266 3261 3181 3166 3095 3059 2973 2961 2897 2887 2825
Intervention 2905 2543 2554 2513 2726 2757 2718 2765 2841 2839 2814 2797 2787 2838 2805 2756 2688 2665 2616 2589 2538 2520 2473 2450 2409
B
105 Mean difference (95% CI): –3·1 (–3·7 to –2·5) at 1 h
Diastolic blood pressure (mm Hg)
–2·4 (–3·0 to –1·8) at 24 h

Figure 2: Physiological 100
measures from random
95
assignment after treatment
to day 7 90
Values are shown for patients
with abnormal systolic blood 85
pressure (A) and abnormal
diastolic blood pressure (B) 80
in 2905 (90·2%) of
75
3221 patients in the care R 15 20 45 1 2 3 4 8 12 16 20 24 2 3 4 5 6 7
bundle group and
Number at risk Minutes Hours Days
3415 (90·7%) of
3767 patients in the usual care Control 3415 2260 2419 2204 2853 2920 2807 3038 3134 3112 3084 3054 3028 3322 3266 3261 3181 3166 3095 3058 2973 2961 2897 2887 2824
group; abnormal glucose (C) in Intervention 2905 2543 2555 2513 2726 2757 2718 2765 2841 2839 2814 2797 2787 2838 2805 2756 2688 2665 2616 2589 2539 2520 2473 2450 2408
1094 (34·5%) of 3175 patients
in the care bundle and C
1319 (37·3%) of 3536 patients 11·0 Mean difference (95% CI): –0·3 (–0·5 to 0·0) at 1 h
in the usual care group; and 10·5 –0·5 (–0·8 to –0·2) at 24 h
abnormal body temperature
Blood glucose (mmol/L)
10·0
(D) in 52 (1·6%) of
3214 patients in the care 9·5
bundle and 68 (1·8%) of 9·0
3716 patients in the usual care
8·5
group. Recordings were at
15-min intervals in the ﬁrst 7·0
hour after commencing 6·5
treatment (time 0), hourly
6·0
from hours 1 to 6, once R 1 2 3 4 8 12 16 20 24 4 8 12 16 20 24 4 8 12 16 20 24
every 6 h until 24 h, and twice Day 1 Day 2 Day 3
per day until day 7. The overall Number at risk
means per treatment group Control 1319 448 414 411 563 540 587 517 557 525 696 597 580 513 551 480 641 564 556 482 524 460
Intervention 1094 626 577 556 625 682 657 646 661 647 708 652 614 621 595 595 626 589 553 561 545 536
and overall difference in
variables between treatment
groups at 1 h and at 24 h were
D
calculated using a repeated- 39·0 Mean difference (95% CI): 0·1 (–0·3 to 0·5) at 1 h
measure linear mixed model 38·6 –0·2 (–0·7 to 0·2) at 24 h
with a fixed effect of 38·2
treatment, a fixed categorical 37·8
Temperature (°C)
effect of time, a fixed 37·4

interaction between 37·0
treatment and time, a 36·6
repeated patient effect (to 36·2
model within-patient 35·8
correlations assuming a 35·4
compound symmetry 35·0
structure), and a random site R 1 2 3 4 8 12 16 20 24 4 8 12 16 20 24 4 8 12 16 20 24
effect, with adjustment for Day 1 Day 2 Day 3
Number at risk
baseline measurements.
Control 68 33 26 26 34 35 33 36 35 35 52 42 38 39 37 33 48 38 33 36 31 29
R indicates time of random Intervention 52 36 36 36 39 38 37 34 35 35 41 36 34 35 31 30 36 33 32 33 32 29
assignment.

Articles
Compared with usual care, implementation of the time- 6 month modified Rankin scale score
sensitive care bundle was also associated with improved 0 1 2 3 4 5 6
survival, health-related quality of life, and serious Intervention 8·5% 31·4% 6·4% 19·0% 11·8% 8·9% 14·1%
adverse events during the follow-up of patients. Given
that 118 of the participating hospitals admitted more Control 10·2% 26·6% 5·9% 17·0% 13·2% 10·1% 17·0%
than 100 patients with acute intracerebral haemorrhage
per year, the size of the beneficial effect equates to an 0 10 20 30 40 50 60 70 80 90 100
Percentage of patients (%)
improved outcome for at least several similar such
patients per site each year. Figure 3: Functional outcome at 90 days in the care bundle and usual care groups, according to scores on the
To our knowledge, this is the first phase 3 multicentre mRS
randomised controlled trial to show a positive outcome Raw distribution of scores on the mRS at 90 days. Scores on the mRS range from 0 to 6, with 0 indicating no
symptoms, 1 indicating symptoms without clinically significant disability, 2 indicating slight disability, 3 indicating
for an acute treatment of intracerebral haemorrhage, and moderate disability, 4 indicating moderately severe disability, 5 indicating severe disability, and 6 indicating death.
one of few trials that have used a stepped wedge, cluster There was a significant difference between the care bundle group and usual care group in the overall distribution of
randomised design across multiple countries.36 Trials in scores (common odds ratio, indicating a lower odds of worse global function outcome on the mRS, 0·86 [95% CI
intracerebral haemorrhage have been complicated by 0·76–0·97]; p=0·015). mRS=modified Rankin Scale.
uncertainty over the optimal timing of the initiation of
interventions and in the assessment of outcomes, and in strong supporting evidence, is applicable to a wide variety
controlling for confounding from comorbid conditions of settings, and has a plausible mechanism of action in
that are either evident at the time of enrolment or arise as that attenuating the growth of the intracerebral
a complication of the intervention or condition of the haemorrhage will reduce the primary injury of dissection
patient during follow-up.13,14,16 Previous trials of early and compression of brain tissue. Although the rapid
intensive blood pressure lowering in acute intracerebral control of blood pressure within an early 6 h inclusion
haemorrhage have produced mixed results, because of time window was emphasised to investigators in this
the variability in the approaches taken towards this study, most of the growth in intracerebral haemorrhage
treatment and the use of small sample sizes. Individual occurs within a few hours of the onset of symptoms.37
patient data meta-analyses, where the reliability of the Moreover, the effect of blood pressure lowering treatment
treatment effect is strengthened by an increase in sample on this intermediate endpoint is modest and does not have
size, the inclusion of a broader range of patients, and a clear time-dependent association defined to date.9 The
adjustment for baseline imbalances in prognostic early correction of an abnormal anticoagulation is likely to
variables, have shown that a careful, targeted, and benefit patients from a similar mechanism of effect,38 but
sustained reduction in systolic blood pressure is safe and in contrast to ageing populations with ready access to
is associated with a better functional outcome after antithrombotic therapies for atrial fibrillation and other
intracerebral haemorrhage.8,9 However, because these cardiovascular disease in high-income countries, only a
data were derived from conventional trials with restricted few patients in our study population presented with
inclusion and exclusion criteria, the results are limited to anticoagulation-associated intracerebral haemorrhage.
patients with a generally good prognosis with intracerebral Neuroinflammation from the breakdown of red blood cells
haemorrhage of a mild-to-moderate severity according to and blood products is a crucial factor in the cause of
standard clinical and imaging criteria. We hypothesised secondary brain injury after intracerebral haemorrhage,
that combining early blood pressure lowering treatment arising in the perihaematoma region and persisting for
with other simple medical interventions would have several days to weeks.39 Because hyperglycaemia40 and
additive benefits in intracerebral haemorrhage. Such a pyrexia41 might potentiate this process, their treatment
multifaceted approach is supported by the Quality in could provide a crucial adjuvant therapeutic strategy.42
Acute Stroke Care trial,17 where improved outcomes were Once again, these exposures occurred in only a few
seen in patients with acute stroke after nurse-initiated participants, with small differences in the amounts of
protocols to manage fever, hyperglycaemia, and swallow control being achieved between randomly assigned
ing dysfunction were implemented at 19 Australian acute groups. Thus, although the overall treatment effect seems
stroke units from 2005 to 2010. Moreover, the to have been driven by intensive blood pressure lowering,
implementation of a quality improvement protocol the active multifaceted management of physiological
involving the reversal of anticoagulation, intensive blood variables and the associated, more intensive, monitoring
pressure lowering, and rapid triage to neurosurgery and and nursing care could have affected behaviours and
critical care improved survival for patients with acute attitudes, including self-management in patients, which
intracerebral haemorrhage at a large tertiary hospital in translated into benefits that extended after discharge from
the UK.18 hospital. A mediation analysis is planned to identify the
However, complex relationships are likely to exist effects of various components of the care bundle and other
between the fidelity of implementation and the effect of key aspects of stroke care on outcomes.
the intervention. Intensive blood pressure lowering was The strengths of our trial include that it comprised an
the central component of the care bundle, because it has evaluation of a complex intervention, for both

Articles
n Odds ratio or mean p value† Intraclass

difference (95% CI)* correlation
coefficient
Primary outcome
Ordinal analysis of category scores measured by the mRS*
Unadjusted 6255 0·86 (0·76 to 0·97)† 0·015 0·05
Adjusted 6069 0·84 (0·73 to 0·97)‡ 0·017 0·07
Secondary outcome
Ordinal analysis of category scores for neurological 6809 0·89 (0·77 to 1·03)†§ 0·12 0·23
impairment or death measured by the NIHSS by day 7 of
treatment
Death or disability at 6 months (mRS scores 3–6) 6255 0·89 (0·78 to 1·02)† 0·10 0·04
Death at 6 months 6440 0·77 (0·63 to 0·95)† 0·015 0·07
Major disability in survivors at 6 months (mRS scores 3–5) 5277 0·96 (0·83 to 1·11)† 0·56 0·03
Health-related quality of life measured by the EQ-5D-3L¶
Mobility 5277 0·87 (0·75 to 1·01)† 0·064 0·02
Self-care 5277 0·93 (0·82 to 1·07)† 0·32 0·02
Usual activities 5277 0·97 (0·85 to 1·10)† 0·62 0·02
Pain or discomfort 5277 0·78 (0·67 to 0·91)† 0·0016 0·02
Anxiety or depression 5277 0·83 (0·69 to 1·00)† 0·046 0·04
Visual analogue scale 5276 1·24 (–0·20 to 2·67)† 0·091 0·02
Mean overall health utility score 6255 0·04 (0·02 to 0·07)† 0·0008 0·06
Hospital discharge by day 7 of treatment 7016 0·72 (0·53 to 0·98)† 0·034 0·33
Residence at home at 6 months 5277 0·94 (0·65 to 1·35)† 0·73 0·03
EQ-5D-3L=EuroQoL Group 5-Dimension Self-Report Questionnaire. mRS=modified Rankin scale. NIHSS=National Institutes of Health Stroke Scale. *The mRS evaluates
global disability; scores range from 0 (no symptoms) to 6 (death). A score of 2–5 indicates some degree of disability. The common odds was estimated from an ordinal
logistic regression model and indicates the odds of worse functional outcome for the care bundle group compared with the usual care group. †Estimates from a logistic or
linear regression model with a random effect of cluster (hospital site) and group assignment (care bundle or usual care) as a fixed effect, and calendar time (6-month
window) as a fixed categorical effect. ‡Adjusted for country (grouped as China vs India, Pakistan, Sri Lanka, and Viet Nam vs Brazil, Peru, Chile, Mexico, and Nigeria),
pre-stroke mRS score, age, sex, baseline NIHSS score, random effect for cluster (hospital site), and group assignment (care bundle or usual care) as a fixed effect, and calendar
time (6-month window) as a fixed categorical effect. §The common odds were estimated from an ordinal logistic regression model and indicate the odds of worse
neurological deterioration measured on the NIHSS (separated into seven categories: <5, 5–9, 10–14, 15–19, 20–24, and ≥25) or death for the care bundle group compared
with the usual care group. Scores on the NIHSS range from 0 to 42, with higher scores indicating more severe neurological deficits. ¶The EQ-5D-3L covers five domains of
health-related quality of life: mobility, self-care, usual activities, pain or discomfort, and anxiety or depression. Each domain has three graded levels of response: no problems,
moderate problems, or extreme problems. Scores are combined to provide an overall health utility score that was calculated with population normal values using China
normative data for China and UK normative data for countries other than China.
Table 3: Clinical outcomes at 6 months
implementation in real-world settings and clinical effects, processes and medication supply happened more widely,
on a robust patient-centred outcome assessed at 6 months. is under investigation through our embedded process
The large sample size allowed for the detection of a evaluation. Given the broad requirement for the
modest, but still clinically worthwhile, benefit that is standardisation of best practice and quality requirements
generalisable, because an ethnically and sociodemo of well organised systems to have little unwarranted
graphically diverse population was included from clinical variation in health care, and the absence of any
different resource settings. The broad inclusion criteria other clearly proven treatment for intracerebral
allowed testing related to the usual triage and care of haemorrhage, we anticipate our findings will also support
patients with acute intracerebral haemorrhage, and the the implementation and enhancement of protocols for
analyses followed a prespecified analysis plan. The intracerebral haemorrhage in high-income countries.
provision of the protocol, management algorithms, and a Our trial has limitations. We chose a stepped wedge
short training package, with individual patient cluster randomised design to allow large numbers of
management left to the discretion of the treating team patients to be rapidly recruited within routine practice,
and supported by performance monitoring, reflects how including those with severe illness and requiring urgent
this intervention might be adopted more widely. Whether neurosurgery, in whom it can be challenging to obtain
the implementation difficulties already noted to have early consent. The introduction of the care bundle
happened in China,24 in relation to concerns about necessitated a change in practice across different clinical
whether the protocol-defined target concentrations for services involved in the flow of patients in hospital. These
blood pressure and glycaemic control might harm issues made individual randomisation unfeasible,
patients, and whether contextual factors related to staffing inequitable, and prone to contamination. However, an

Articles
important potential limitation of this study design is Declaration of interests

confounding effects from secular trends by calendar time, LS reports funding from the Medical Research Council of the UK,
Sichuan Credit Pharmaceutic, and Takeda China; and speaker fees from
related to the sequential implementation of the
Takeda China. CSA has received grants from the National Health and
intervention across participating sites, and from Medical Research Council and Medical Research Futures Fund of
disruption to the presentation and management of Australia, the Medical Research Council of the UK, Penumbra, and
patients created by the COVID-19 pandemic. There was Takeda China; is also the chair of the data and safety monitoring boards
for several trials; is a board member of WHO; and is the Editor-in-Chief
considerable heterogeneity in the recruitment of patients of Cerebrovascular Disease. CY has received funding from West China
across sites and in the timing of the phases of the trial. Hospital. All other authors declare no competing interests.
Recruitment was also extended for patients in the Data sharing
intervention group towards the end of the trial, which Individual, de-identified participant data used in these analyses will be
might have introduced a degree of selection bias. To shared on request from any qualified investigator after the approval of a
accommodate this heterogeneity, the extent of which was protocol and signed data access agreement via both the trial steering
committee and the research office of The George Institute for Global
not fully appreciated until the unmasking of the data, we Health (Sydney, NSW, Australia).
made a post-hoc decision to adjust the analyses according
Acknowledgments
to calendar time instead of by trial period, as recommended The study is supported by an award (grant reference number
in another stepped wedge trial with delays in cluster MR/T005009/1) jointly funded by the Department of Health and Social
recruitment.31 We could not find any substantial Care, the Foreign, Commonwealth & Development Office, the Medical
differences in the characteristics of patients or their Research Council, and the Wellcome Trust (all London, UK);
the West China Hospital Outstanding Discipline Development 1–3-5
management by randomised group, overall, and by programme (number ZY2016102); National Health and Medical Research
region, which is especially relevant for the use of Council of Australia (number APP1149987); Sichuan Credit
decompressive surgery, mechanical ventilation, and early Pharmaceutical; and Takeda (China). We thank the independent
withdrawal of care, because these interventions have the members and patient representatives of the trial steering committee:
Thompson Robinson (Chair, University of Leicester, Leicester, UK),
most potential to influence survival from intracerebral J Jaime Miranda (Deputy Chair, Centro de Excelencia en Enfermedades
haemorrhage. Even so, the possibility of unmeasured Crónicas, Universidad Peruana Cayetano Heredia, Lima, Peru),
confounding from other changes in the management of Adrian Parry-Jones (University of Manchester, Manchester, UK),
Nikola Sprigg (University of Nottingham, Nottingham, UK),
patients over time cannot be completely excluded. Our
Ann Bamford (Manchester, UK), Olivia Smith (Nottingham, UK),
trial was not powered to detect differences across and Caroline Harris (Global Health Strategy, Medical Research Council,
secondary outcomes, with statistical significance varying London, UK). We acknowledge the EuroQol Group for use of the
across the secondary outcomes and generally not present EQ-5D-3L. CSA is a senior investigator fellow for the National Health and
Medical Research Council of Australia. We thank the investigators and
after adjustment for patient characteristics. However, the
research staff at the participating sites of the various participating
direction of the effect was consistently in favour of the emergency departments, intensive care units, and stroke units, and
care bundle across all secondary outcomes. There were neurosurgery and neurology departments (appendix pp 4–9),
logistical and practical issues in delivering the and executive staff at The George Institute for Global Health China for
their support of the study. We also thank the participants, their relatives,
intervention, following up patients, and in monitoring the
and their families.
data quality in LMICs. Bias and measurement error are
References
likely to have influenced the assessment of baseline and 1 Cordonnier C, Demchuk A, Ziai W, Anderson CS. Intracerebral
outcome measures, such as in the primary use of proxies haemorrhage: current approaches to acute management. Lancet
with wide-ranging sociodemographic and cultural 2018; 392: 1257–68.
2 GBD 2019 Stroke Collaborators. Global, regional, and national
backgrounds to assess the health-related quality of life in burden of stroke and its risk factors, 1990–2019: a systematic
patients.43 analysis for the Global Burden of Disease Study 2019. Lancet Neurol
In summary, the findings of our trial provide evidence to 2021; 20: 795–820.
3 Greenberg SM, Ziai WC, Cordonnier C, et al. 2022 Guideline for
support the adoption of an active protocol for intensive the management of patients with spontaneous intracerebral
blood pressure lowering and the associated management of hemorrhage: a guideline from the American Heart Association/
key abnormal physiological variables within several hours American Stroke Association. Stroke 2022; 53: e282–361.
after the onset of signs to improve the recovery 4 Sandset EC, Anderson CS, Bath PM, et al. European Stroke
Organisation (ESO) guidelines on blood pressure management in
of patients presenting with acute intracerebral haemorrhage. acute ischaemic stroke and intracerebral haemorrhage. Eur Stroke J
Contributors 2021; 6: II.
CSA, LS, CY, LM, XH, and XC designed the study. LS, XC, and MO 5 Stroke Foundation. Living Clinical Guidelines for Stroke
provided quality control oversight. QL did statistical analyses and reports. Management. Chapter 3 Acute medical and surgical management.
https://informme.org.au/en/Guidelines/Living-Clinical-Guidelines-
LB wrote the statistical analysis plan with input from CSA and LS.
for-Stroke-Management (accessed Jan 5, 2023).
PM-V, AdS, NHT, KWW, JDP, MW, OMP-N, CA, and AA provided
6 Cao Y, Yu S, Zhang Q, et al. Chinese Stroke Association guidelines
comments on the study design and were responsible for data collection
for clinical management of cerebrovascular disorders: executive
and quality control procedures. LS and CSA wrote the first draft of the summary and 2019 update of clinical management of intracerebral
manuscript. YS, SY, QH, BC, YuL, and EC read the imaging scans. haemorrhage. Stroke Vasc Neurol 2020; 5: 396–402.
LM, XH, QL, LB, CSA, CY, LS, XC, and JX had full access to all the study 7 Shoamanesh A, Lindsay MP, Castellucci LA, et al. Canadian stroke
data and verified the data, and they had final responsibility for the best practice recommendations: management of spontaneous
decision to submit the paper for publication. All authors commented on intracerebral hemorrhage, 7th edition update 2020. Int J Stroke
drafts of the manuscript and approved the final manuscript for 2021; 16: 321–41.
submission.

Articles
8 Moullaali TJ, Wang X, Martin RH, et al. Blood pressure control and 25 Anderson CS, Arima H, Lavados P, et al. Cluster-randomised,
clinical outcomes in acute intracerebral haemorrhage: a pre- crossover trial of head positioning in acute stroke. N Engl J Med
planned pooled analysis of individual participant data. Lancet Neurol 2017; 376: 2437–47.
2019; 18: 857–64. 26 Anderson CS, Heeley E, Huang Y, et al. Rapid blood-pressure
9 Moullaali TJ, Wang X, Sandset EC, et al. Early lowering of blood lowering in patients with acute intracerebral hemorrhage.
pressure after acute intracerebral haemorrhage: a systematic review N Engl J Med 2013; 368: 2355–65.
and meta-analysis of individual patient data. 27 Whitehead J. Sample size calculations for ordered categorical data.
J Neurol Neurosurg Psychiatry 2022; 93: 6–13. Stat Med 1993; 12: 2257–71.
10 Zheng J, Li H, Lin S, et al. Perioperative antihypertensive treatment 28 Woertman W, de Hoop E, Moerbeek M, Zuidema SU, Gerritsen DL,
in patients with spontaneous intracerebral hemorrhage. Stroke 2017; Teerenstra S. Stepped wedge designs could reduce the required
48: 216–18. sample size in cluster randomized trials. J Clin Epidemiol 2013;
11 Parry-Jones AR, Paley L, Bray BD, et al. Care-limiting decisions in 66: 752–58.
acute stroke and association with survival: analyses of UK national 29 Hussey MA, Hughes JP. Design and analysis of stepped wedge
quality register data. Int J Stroke 2016; 11: 321–31. cluster randomized trials. Contemp Clin Trials 2007; 28: 182–91.
12 Morgenstern LB, Zahuranec DB, Sánchez BN, et al. Full medical 30 Kasza J, Bowden R, Hooper R, Forbes AB. The batched stepped
support for intracerebral hemorrhage. Neurology 2015; 84: 1739–44. wedge design: a design robust to delays in cluster recruitment.
13 The Hemorrhagic Stroke Academia Industry (HEADS) Roundtable Stat Med 2022; 41: 3627–41.
Participants, The Second HEADS Roundtable Participants. 31 Li F, Hughes JP, Hemming K, Taljaard M, Melnick ER,
Recommendations for clinical trials in ICH: the second Heagerty PJ. Mixed-effects models for the design and analysis of
hemorrhagic stroke academia industry roundtable. Stroke 2020; stepped wedge cluster randomized trials: an overview.
51: 1333–38. Stat Methods Med Res 2021; 30: 612–39.
14 Broderick JP, Grotta JC, Naidech AM, et al. The story of 32 Hemming K, Taljaard M, Forbes A. Analysis of cluster randomised
intracerebral hemorrhage from recalcitrant to treatable disease. stepped wedge trials with repeated cross-sectional samples. Trials
Stroke 2021; 52: 1905–14. 2017; 18: 101.
15 Morotti A, Arba F, Boulouis G, Charidimou A. Noncontrast CT 33 Haybittle JL. Repeated assessment of results in clinical trials of
markers of intracerebral hemorrhage expansion and poor outcome: cancer treatment. Br J Radiol 1971; 44: 793–97.
a meta-analysis. Neurology 2020; 95: 632–43. 34 Abdi H. Holm’s sequential Bonferroni procedure. In: Salkind N, ed.
16 Dammers R, Beck J, Volovici V, Anderson CS, Klijn CJM. Encyclopedia of research design. Thousand Oaks, CA: Sage, 2010,
Advancing the surgical treatment of intracerebral hemorrhage: 574–89.
study design and research directions. World Neurosurg 2022; 35 Pocock SJ, Rossello X, Owen R, Collier TJ, Stone GW, Rockhold FW.
161: 367–75. Primary and secondary outcome reporting in randomized trials:
17 Middleton S, McElduff P, Ward J, et al. Implementation of JACC state-of the-art review. J Am Coll Cardiol 2021; 78: 827–39.
evidence-based treatment protocols to manage fever, 36 Vousden N, Lawley E, Nathan HL, et al. Effect of a novel vital sign
hyperglycaemia, and swallowing dysfunction in acute stroke device on maternal mortality and morbidity in low-resource
(QASC): a cluster randomised controlled trial. Lancet 2011; settings: a pragmatic, stepped-wedge, cluster-randomised controlled
378: 1699–706. trial. Lancet Glob Health 2019; 7: e347–56.
18 Parry-Jones AR, Sammut-Powell C, Paroutoglou K, et al. 37 Al-Shahi Salman R, Frantzias J, Lee RJ, et al. Absolute risk and
Intracerebral hemorrhage care bundle is associated with lower case predictors of the growth of acute spontaneous intracerebral
fatality. Ann Neurol 2019; 86: 495–503. haemorrhage: a systematic review and meta-analysis of individual
19 Song L, Wang X, Ouyang M, et al. Associations of an abnormal patient data. Lancet Neurol 2018; 17: 885–94.
physiological score with outcomes in acute intracerebral 38 Steiner T, Poli S, Griebe M, et al. Fresh frozen plasma versus
hemorrhage: INTERACT2 study. Stroke 2021; 52: 722–25. prothrombin complex concentrate in patients with intracranial
20 Hemming K, Taljaard M, McKenzie JE, et al. Reporting of stepped haemorrhage related to vitamin K antagonists (INCH):
wedge cluster randomised trials: extension of the CONSORT 2010 a randomised trial. Lancet Neurol 2016; 15: 566–73.
statement with explanation and elaboration. BMJ 2018; 363: k1614. 39 Xi G, Keep RF, Hoff JT. Mechanisms of brain injury after
21 Song L, Hu X, Ma L, et al. INTEnsive care bundle with blood intracerebral haemorrhage. Lancet Neurol 2006; 5: 53–63.
pressure reduction in acute cerebral hemorrhage trial 40 Rom S, Zuluaga-Ramirez V, Gajghate S, et al. Hyperglycemia-driven
(INTERACT3): study protocol for a pragmatic stepped-wedge neuroinflammation compromises BBB leading to memory loss in
cluster-randomized controlled trial. Trials 2021; 22: 943. both diabetes mellitus (DM) type 1 and type 2 mouse models.
22 Billot L, Song L, Hu X, et al. Statistical analysis plan for the Mol Neurobiol 2019; 56: 1883–96.
INTEnsive care bundle with blood pressure reduction in acute 41 Birg T, Ortolano F, Wiegers EJA, et al. Brain temperature influences
cerebral hemorrhage trial: a stepped-wedge cluster randomized intracranial pressure and cerebral perfusion pressure after
controlled trial. Cerebrovasc Dis 2022; published online Sept 5. traumatic brain injury: a CENTER-TBI study. Neurocrit Care 2021;
https://doi.org/10.1159/000526384. 35: 651–61.
23 American Diabetes Association. Standards of medical care in 42 Kearns KN, Ironside N, Park MS, et al. Neuroprotective therapies
diabetes–2015: summary of revisions. Diabetes Care 2015; for spontaneous intracerebral hemorrhage. Neurocrit Care 2021;
38 (suppl): S4. 35: 862–86.
24 Ouyang M, Anderson CS, Song L, et al. Implementing a goal- 43 Williams LS, Bakas T, Brizendine E, et al. How valid are family
directed care bundle after acute intracerebral haemorrhage: process proxy assessments of stroke patients’ health-related quality of life?
evaluation for the Third Intensive Care Bundle with Blood Pressure Stroke 2006; 37: 2081–85.
Reduction in Acute Cerebral Haemorrhage trial study in China.
Cerebrovasc Dis 2022; 51: 373–83.

Pi Is 0140673623008061

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Pi Is 0140673623008061

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pi Is 0140673623008061

Uploaded by

Copyright:

Available Formats

Articles

The third Intensive Care Bundle with Blood Pressure

www.thelancet.com Vol 402 July 1, 2023 27

28 www.thelancet.com Vol 402 July 1, 2023

www.thelancet.com Vol 402 July 1, 2023 29

30 www.thelancet.com Vol 402 July 1, 2023

www.thelancet.com Vol 402 July 1, 2023 31

treatment-by-time interactions fixed effects, site as a

32 www.thelancet.com Vol 402 July 1, 2023

107 (3·3%) patients in the care bundle group had a

www.thelancet.com Vol 402 July 1, 2023 33

(median, 1·0 h [IQR 1·0–1·0] vs 1·0 h [1·0–1·0]). Patients

34 www.thelancet.com Vol 402 July 1, 2023

www.thelancet.com Vol 402 July 1, 2023 35

–2·4 (–3·0 to –1·8) at 24 h

effect of time, a fixed 37·4

36 www.thelancet.com Vol 402 July 1, 2023

www.thelancet.com Vol 402 July 1, 2023 37

n Odds ratio or mean p value† Intraclass

Table 3: Clinical outcomes at 6 months

38 www.thelancet.com Vol 402 July 1, 2023

important potential limitation of this study design is Declaration of interests

www.thelancet.com Vol 402 July 1, 2023 39

40 www.thelancet.com Vol 402 July 1, 2023

You might also like