kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Clustering identifies endotypes of traumatic brain injury in an intensive care cohort: a CENTER-TBI study
KTH, School of Electrical Engineering and Computer Science (EECS), Computer Science, Computational Science and Technology (CST). Karolinska Institutet, Stockholm, Sweden.ORCID iD: 0000-0003-4918-1482
KTH, School of Electrical Engineering and Computer Science (EECS), Computer Science, Computational Science and Technology (CST).ORCID iD: 0000-0001-8577-6745
Univ Milan, Fdn IRCCS Ca Granda Ospeda Maggiore Policlin, Neurosci Intens Care Unit, Dept Pathophysiol & Transplants, Milan, Italy..
Leiden Univ, Dept Biomed Data Sci, Med Ctr, Leiden, Netherlands..
Show others and affiliations
2022 (English)In: Critical Care, ISSN 1364-8535, E-ISSN 1466-609X, Vol. 26, no 1, article id 228Article in journal (Refereed) Published
Abstract [en]

Background: While the Glasgow coma scale (GCS) is one of the strongest outcome predictors, the current classification of traumatic brain injury (TBI) as'mild" 'moderate'or'severe' based on this fails to capture enormous heterogeneity in pathophysiology and treatment response. We hypothesized that data-driven characterization of TBl could identify distinct endotypes and give mechanistic insights. Methods: We developed an unsupervised statistical clustering model based on a mixture of probabilistic graphs for presentation (<24 h) demographic, clinical, physiological, laboratory and imaging data to identify subgroups of TBl patients admitted to the intensive care unit in the CENTER-TBI dataset (N= 1,728). A cluster similarity index was used for robust determination of optimal cluster number. Mutual information was used to quantify feature importance and for cluster interpretation. Results: Six stable endotypes were identified with distinct GCS and composite systemic metabolic stress profiles, distinguished by GCS, blood lactate, oxygen saturation, serum creatinine, glucose, base excess, pH, arterial partial pressure of carbon dioxide, and body temperature. Notably, a cluster with 'moderate'TBI (by traditional classification) and deranged metabolic profile, had a worse outcome than a cluster with 'severe'GCS and a normal metabolic profile. Addition of cluster labels significantly improved the prognostic precision of the IMPACT (International Mission for Prognosis and Analysis of Clinical trials in TBI) extended model, for prediction of both unfavourable outcome and mortality (both p <0.001). Conclusions: Six stable and clinically distinct TBI endotypes were identified by probabilistic unsupervised clustering. In addition to presenting neurology, a profile of biochemical derangement was found to be an important distinguishing feature that was both biologically plausible and associated with outcome. Our work motivates refining current TBI classifications with factors describing metabolic stress. Such data-driven clusters suggest TBI endotypes that merit investigation to identify bespoke treatment strategies to improve care.

Place, publisher, year, edition, pages
Springer Nature , 2022. Vol. 26, no 1, article id 228
Keywords [en]
Traumatic brain injury, Endotypes, Intensive care unit, Critical care, Unsupervised clustering, Machine learning
National Category
Nursing
Identifiers
URN: urn:nbn:se:kth:diva-316301DOI: 10.1186/s13054-022-04079-wISI: 000831208500002PubMedID: 35897070Scopus ID: 2-s2.0-85135370588OAI: oai:DiVA.org:kth-316301DiVA, id: diva2:1686858
Note

QC 20220811

Available from: 2022-08-11 Created: 2022-08-11 Last updated: 2023-05-04Bibliographically approved
In thesis
1. Pathophysiological characterization of traumatic brain injury using novel analytical methods
Open this publication in new window or tab >>Pathophysiological characterization of traumatic brain injury using novel analytical methods
2023 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Severity of traumatic brain injury is usually classified by Glasgow coma scale (GCS) as “mild”, "moderate" or "severe’, which does not capture the heterogeneity of the disease. According to current guidelines, intracranial pressure (ICP) should not exceed 22 mmHg, with no further recommendations concerning individualization or tolerable duration of intracranial hypertension. The aims of this thesis were to identify subgroups of patients beyond characterization using GCS, and to investigate the impact of duration and magnitude of intracranial hypertension on outcome, using data from the observational prospective study Collaborative European neurotrauma effectiveness research in TBI (CENTER-TBI). 

To investigate the temporal aspect of tolerable ICP elevations, we examined the correlation between dose of ICP and outcome represented by 6-month Glasgow outcome scale extended (GOSE). ICP dose was represented both by the number of events above thresholds for ICP magnitude and duration and by area under the ICP curve (i.e., “pressure time dose” (PTD)). A variation in tolerable ICP thresholds of 18 mmHg +/- 4 mmHg (2 standard deviations (SD)) for events with duration longer than five minutes was identified using a bootstrapping technique. PTD was correlated to both mortality and unfavorable outcome. 

A cerebrovascular autoregulation (CA) dependent ICP tolerability was identified. If CA was impaired, no tolerable ICP magnitude and duration thresholds were identified, while if CA was intact, both 19 mmHg for 5 minutes or longer and 15 mmHg for 50 minutes or longer were correlated to worse outcome. While no significant difference in PTD was seen between favorable and unfavorable outcome if CA was intact, there was a significant difference if CA was impaired. In a multivariable analysis, PTD did not remain a significant predictor of outcome when adjusting for other known predictors in TBI. In a causal inference analysis, both cerebrovascular autoregulation status and ICP-lowering therapies represented by the therapy intensity level (TIL) have a directional relationship with outcome. However, no direct causal relationship of ICP towards outcome was found. 

By applying an unsupervised clustering method, we identified six distinct admission clusters defined by GCS, lactate, oxygen saturation (SpO2), creatinine, glucose, base excess, pH, PaCO2, and body temperature. These clusters can be summarized in clinical presentation and metabolic profile. When clustering longitudinal features during the first week in the intensive care unit (ICU), no optimal number of clusters could be seen. However, glucose variation, a panel of brain biomarkers, and creatinine consistently described trajectories. Although no information on outcome was included in the models, both admission clusters and trajectories showed clear outcome differences, with mortality from 7 to 40% in the admission clusters and 4 to 85% in the trajectories. Adding cluster or trajectory labels to the established outcome prediction IMPACT model significantly improved outcome predictions. 

The results in this thesis support the importance of cerebrovascular autoregulation status as it was found that CA status was more informative towards outcome than ICP magnitude and duration. There was a variation in tolerable ICP intensity and duration dependent on whether CA was intact. Distinct clusters defined by GCS and metabolic profiles related to outcome suggest the importance of an extracranial evaluation in addition to GCS in TBI patients. Longitudinal trajectories of TBI patients in the ICU are highly characterized by glucose variation, brain biomarkers and creatinine.

Place, publisher, year, edition, pages
Stockholm: Karolinska Institutet, 2023. p. 67
Keywords
Traumatic brain injury; Intracranial pressure; clustering
National Category
Neurology Medical and Health Sciences Clinical Medicine
Research subject
Medical Technology
Identifiers
urn:nbn:se:kth:diva-326337 (URN)978-91-8016-973-8 (ISBN)
Public defence
2023-05-26, Torsten Gordh Auditorium, S2:02, Norrbacka, Karolinska Universitetssjukhuset, Solna, 09:00 (English)
Opponent
Supervisors
Note

QC 20230503

Available from: 2023-05-04 Created: 2023-05-02 Last updated: 2023-08-29Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textPubMedScopus

Authority records

Åkerlund, CeciliaHolst, Anders

Search in DiVA

By author/editor
Åkerlund, CeciliaHolst, AndersErcole, Ari
By organisation
Computational Science and Technology (CST)
In the same journal
Critical Care
Nursing

Search outside of DiVA

GoogleGoogle Scholar

doi
pubmed
urn-nbn

Altmetric score

doi
pubmed
urn-nbn
Total: 225 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf