Clustering identifies endotypes of traumatic brain injury in an intensive care cohort: a CENTER-TBI study.
Critical care
Endotypes
Intensive care unit
Machine learning
Traumatic brain injury
Unsupervised clustering
Journal
Critical care (London, England)
ISSN: 1466-609X
Titre abrégé: Crit Care
Pays: England
ID NLM: 9801902
Informations de publication
Date de publication:
27 07 2022
27 07 2022
Historique:
received:
20
02
2022
accepted:
02
07
2022
entrez:
27
7
2022
pubmed:
28
7
2022
medline:
30
7
2022
Statut:
epublish
Résumé
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 TBI could identify distinct endotypes and give mechanistic insights. 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 TBI 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. 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). 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. Trial registration The core study was registered with ClinicalTrials.gov, number NCT02210221 , registered on August 06, 2014, with Resource Identification Portal (RRID: SCR_015582).
Sections du résumé
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 TBI 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 TBI 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. Trial registration The core study was registered with ClinicalTrials.gov, number NCT02210221 , registered on August 06, 2014, with Resource Identification Portal (RRID: SCR_015582).
Identifiants
pubmed: 35897070
doi: 10.1186/s13054-022-04079-w
pii: 10.1186/s13054-022-04079-w
pmc: PMC9327174
doi:
Banques de données
ClinicalTrials.gov
['NCT02210221']
Types de publication
Clinical Trial
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
228Investigateurs
Cecilia Åkerlund
(C)
Krisztina Amrein
(K)
Nada Andelic
(N)
Lasse Andreassen
(L)
Audny Anke
(A)
Anna Antoni
(A)
Gérard Audibert
(G)
Philippe Azouvi
(P)
Maria Luisa Azzolini
(ML)
Ronald Bartels
(R)
Pál Barzó
(P)
Romuald Beauvais
(R)
Ronny Beer
(R)
Bo-Michael Bellander
(BM)
Antonio Belli
(A)
Habib Benali
(H)
Maurizio Berardino
(M)
Luigi Beretta
(L)
Morten Blaabjerg
(M)
Peter Bragge
(P)
Alexandra Brazinova
(A)
Vibeke Brinck
(V)
Joanne Brooker
(J)
Camilla Brorsson
(C)
Andras Buki
(A)
Monika Bullinger
(M)
Manuel Cabeleira
(M)
Alessio Caccioppola
(A)
Emiliana Calappi
(E)
Maria Rosa Calvi
(MR)
Peter Cameron
(P)
Guillermo Carbayo Lozano
(GC)
Marco Carbonara
(M)
Simona Cavallo
(S)
Giorgio Chevallard
(G)
Arturo Chieregato
(A)
Giuseppe Citerio
(G)
Hans Clusmann
(H)
Mark Coburn
(M)
Jonathan Coles
(J)
Jamie D Cooper
(JD)
Marta Correia
(M)
Amra Čović
(A)
Nicola Curry
(N)
Endre Czeiter
(E)
Marek Czosnyka
(M)
Claire DahyotFizelier
(C)
Paul Dark
(P)
Helen Dawes
(H)
Véronique De Keyser
(V)
Vincent Degos
(V)
Francesco Della Corte
(FD)
Hugo den Boogert
(HD)
Bart Depreitere
(B)
Đula Đilvesi
(Đ)
Abhishek Dixit
(A)
Emma Donoghue
(E)
Jens Dreier
(J)
GuyLoup Dulière
(G)
Ari Ercole
(A)
Patrick Esser
(P)
Erzsébet Ezer
(E)
Martin Fabricius
(M)
Valery L Feigin
(VL)
Kelly Foks
(K)
Shirin Frisvold
(S)
Alex Furmanov
(A)
Pablo Gagliardo
(P)
Damien Galanaud
(D)
Dashiell Gantner
(D)
Guoyi Gao
(G)
Pradeep George
(P)
Alexandre Ghuysen
(A)
Lelde Giga
(L)
Ben Glocker
(B)
Jagoš Golubovic
(J)
Pedro A Gomez
(PA)
Johannes Gratz
(J)
Benjamin Gravesteijn
(B)
Francesca Grossi
(F)
Russell L Gruen
(RL)
Deepak Gupta
(D)
Juanita A Haagsma
(JA)
Iain Haitsma
(I)
Raimund Helbok
(R)
Eirik Helseth
(E)
Lindsay Horton
(L)
Jilske Huijben
(J)
Peter J Hutchinson
(PJ)
Bram Jacobs
(B)
Stefan Jankowski
(S)
Mike Jarrett
(M)
Jiyao Jiang
(J)
Faye Johnson
(F)
Kelly Jones
(K)
Mladen Karan
(M)
Angelos G Kolias
(AG)
Erwin Kompanje
(E)
Daniel Kondziella
(D)
Evgenios Kornaropoulos
(E)
LarsOwe Koskinen
(L)
Noémi Kovács
(N)
Ana Kowark
(A)
Alfonso Lagares
(A)
Linda Lanyon
(L)
Steven Laureys
(S)
Fiona Lecky
(F)
Didier Ledoux
(D)
Rolf Lefering
(R)
Valerie Legrand
(V)
Aurelie Lejeune
(A)
Leon Levi
(L)
Roger Lightfoot
(R)
Hester Lingsma
(H)
Andrew I R Maas
(AIR)
Ana M CastañoLeón
(AM)
Marc Maegele
(M)
Marek Majdan
(M)
Alex Manara
(A)
Geoffrey Manley
(G)
Costanza Martino
(C)
Hugues Maréchal
(H)
Julia Mattern
(J)
Catherine McMahon
(C)
Béla Melegh
(B)
David Menon
(D)
Tomas Menovsky
(T)
Ana Mikolic
(A)
Benoit Misset
(B)
Visakh Muraleedharan
(V)
Lynnette Murray
(L)
Ancuta Negru
(A)
David Nelson
(D)
Virginia Newcombe
(V)
Daan Nieboer
(D)
József Nyirádi
(J)
Otesile Olubukola
(O)
Matej Oresic
(M)
Fabrizio Ortolano
(F)
Aarno Palotie
(A)
Paul M Parizel
(PM)
JeanFrançois Payen
(J)
Natascha Perera
(N)
Vincent Perlbarg
(V)
Paolo Persona
(P)
Wilco Peul
(W)
Anna Piippo-Karjalainen
(A)
Matti Pirinen
(M)
Dana Pisica
(D)
Horia Ples
(H)
Suzanne Polinder
(S)
Inigo Pomposo
(I)
Jussi P Posti
(JP)
Louis Puybasset
(L)
Andreea Radoi
(A)
Arminas Ragauskas
(A)
Rahul Raj
(R)
Malinka Rambadagalla
(M)
Isabel Retel Helmrich
(IR)
Jonathan Rhodes
(J)
Sylvia Richardson
(S)
Sophie Richter
(S)
Samuli Ripatti
(S)
Saulius Rocka
(S)
Cecilie Roe
(C)
Olav Roise
(O)
Jonathan Rosand
(J)
Jeffrey V Rosenfeld
(JV)
Christina Rosenlund
(C)
Guy Rosenthal
(G)
Rolf Rossaint
(R)
Sandra Rossi
(S)
Daniel Rueckert
(D)
Martin Rusnák
(M)
Juan Sahuquillo
(J)
Oliver Sakowitz
(O)
Renan SanchezPorras
(R)
Janos Sandor
(J)
Nadine Schäfer
(N)
Silke Schmidt
(S)
Herbert Schoechl
(H)
Guus Schoonman
(G)
Rico Frederik Schou
(RF)
Elisabeth Schwendenwein
(E)
Charlie Sewalt
(C)
Ranjit D Singh
(RD)
Toril Skandsen
(T)
Peter Smielewski
(P)
Abayomi Sorinola
(A)
Emmanuel Stamatakis
(E)
Simon Stanworth
(S)
Robert Stevens
(R)
William Stewart
(W)
Ewout W Steyerberg
(EW)
Nino Stocchetti
(N)
Nina Sundström
(N)
Riikka Takala
(R)
Viktória Tamás
(V)
Tomas Tamosuitis
(T)
Mark Steven Taylor
(MS)
Braden Te Ao
(BT)
Olli Tenovuo
(O)
Alice Theadom
(A)
Matt Thomas
(M)
Dick Tibboel
(D)
Marjolein Timmers
(M)
Christos Tolias
(C)
Tony Trapani
(T)
Cristina Maria Tudora
(CM)
Andreas Unterberg
(A)
Peter Vajkoczy
(P)
Shirley Vallance
(S)
Egils Valeinis
(E)
Zoltán Vámos
(Z)
Mathieu van der Jagt
(M)
Gregory Van der Steen
(G)
Joukje van der Naalt
(J)
Jeroen T J M van Dijck
(JTJM)
Inge A van Erp
(IA)
Thomas A van Essen
(TA)
Wim Van Hecke
(W)
Caroline van Heugten
(C)
Dominique Van Praag
(D)
Ernest van Veen
(E)
Thijs Vande Vyvere
(TV)
Roel P J van Wijk
(RPJ)
Alessia Vargiolu
(A)
Emmanuel Vega
(E)
Kimberley Velt
(K)
Jan Verheyden
(J)
Paul M Vespa
(PM)
Anne Vik
(A)
Rimantas Vilcinis
(R)
Victor Volovici
(V)
Nicole von Steinbüchel
(N)
Daphne Voormolen
(D)
Petar Vulekovic
(P)
Kevin K W Wang
(KKW)
Daniel Whitehouse
(D)
Eveline Wiegers
(E)
Guy Williams
(G)
Lindsay Wilson
(L)
Stefan Winzeck
(S)
Stefan Wolf
(S)
Zhihui Yang
(Z)
Peter Ylén
(P)
Alexander Younsi
(A)
Frederick A Zeiler
(FA)
Veronika Zelinkova
(V)
Agate Ziverte
(A)
Tommaso Zoerle
(T)
Informations de copyright
© 2022. The Author(s).
Références
Am J Respir Crit Care Med. 2019 Apr 15;199(8):980-986
pubmed: 30365341
Lancet Neurol. 2019 Oct;18(10):923-934
pubmed: 31526754
Crit Care. 2006;10 Suppl 2:S2
pubmed: 17352796
Intensive Care Med. 2019 Dec;45(12):1783-1794
pubmed: 31659383
PLoS Med. 2008 Aug 5;5(8):e165; discussion e165
pubmed: 18684008
Curr Opin Crit Care. 2021 Apr 1;27(2):80-86
pubmed: 33534257
Nature. 2015 Nov 19;527(7578):S193-7
pubmed: 26580327
Lancet Respir Med. 2014 Aug;2(8):611-20
pubmed: 24853585
N Engl J Med. 2012 Dec 27;367(26):2471-81
pubmed: 23234472
Intensive Care Med. 2020 May;46(5):995-1004
pubmed: 32100061
Nat Rev Neurol. 2013 Apr;9(4):231-6
pubmed: 23443846
Neurosurgery. 2005 Dec;57(6):1173-82; discussion 1173-82
pubmed: 16331165
Neurosurgery. 2017 Jan 01;80(1):6-15
pubmed: 27654000
Neurosurgery. 2014 Dec;75(6):632-46; discussion 646-7
pubmed: 25181434
Neurosurgery. 2015 Jan;76(1):67-80
pubmed: 25525693
J Clin Med. 2021 Jan 04;10(1):
pubmed: 33406786
J Neurotrauma. 2020 Apr 1;37(7):1002-1010
pubmed: 31672086
Am J Respir Crit Care Med. 2016 Jul 15;194(2):147-55
pubmed: 27244481
J Neurotrauma. 2010 Jan;27(1):51-64
pubmed: 19698072
Lancet Neurol. 2017 Dec;16(12):987-1048
pubmed: 29122524
Clin Rev Allergy Immunol. 2019 Apr;56(2):219-233
pubmed: 30206782
J Neurotrauma. 2016 Aug 15;33(16):1461-78
pubmed: 26711675
JAMA Neurol. 2021 Sep 1;78(9):1137-1148
pubmed: 34279565
Neural Comput. 2004 Jun;16(6):1299-323
pubmed: 15130251
Crit Care. 2014 Apr 03;18(2):R60
pubmed: 24708781
Lancet Neurol. 2010 May;9(5):543-54
pubmed: 20398861
PLoS Med. 2017 Aug 3;14(8):e1002368
pubmed: 28771476
J Neurotrauma. 2020 Jun 15;37(12):1431-1444
pubmed: 32008422