sÍndrome de insuficiencia respiratoria aguda:...

SÍNDROME DE INSUFICIENCIA RESPIRATORIA AGUDA: BIOMARCADORES

TTE NAV SSN (N) MC CARLOS ALBERTO PEÑA PÉREZ MEDICINA INTERNA / MEDICINA DEL ENFERMO EN ESTADO CRÍTICO JEFATURA UNIDAD DE TERAPIA INTENSIVA HOSPITAL GENERAL NAVAL DE ALTA ESPECIALIDAD MÉDICO ADSCRITO UNIDAD DE TERAPIA INTENSIVA HOSPITAL MÉDICA SUR

•  Los Biomarcadores han surgido como ejemplares protagonistas de lamedicinatraslacional…

• Muchoshansidoevaluadosparaelreconocimientooportuno,tratamientotempranoyseguimientodeunavariedaddepatologías...

•  La sensibilidadde los biomarcadores hamejorado considerablemente enlosúl>mosaños,sinembargo,laespecificidadsiguesiendopobre...

•  La inves>gación de los biomarcadores >ene un enorme potencial dediagnós>coypronós>co...

Honore PM, JacobsR,Hendrickx I, et al.Biomarkers in cri-cal illness: havewemadeprogress? Interna'onal Journal ofNephrology andRenovascularDisease2016;9:253-256.

•  ElSIRAesunacomplicaciónpotencialmentemortaldeenfermedadesagudas tales como sepsis, neumonía, trauma, pancrea>>s aguda,aspiracióndelcontenidogástricoyelcasiahogamiento…

•  El SIRA se caracteriza por lesión alveolar difusa, edema pulmonar,inflamaciónderivadadelosneutrófilosydisfuncióndelsurfactante...

•  Las manifestaciones clínicas principales son el descenso en ladistensibilidadpulmonar,hipoxemiaengradosvariableseinfiltradospulmonaresbilaterales...

Honore PM, JacobsR,Hendrickx I, et al.Biomarkers in cri-cal illness: havewemadeprogress? Interna'onal Journal ofNephrology andRenovascularDisease2016;9:253-256.

•  Un biomarcador ideal debe indicar una relación clara con el eventofisiopatológico, debe ser confiable, reproducible, específico de laenfermedad y sensible, debe ser muestreado por métodos simples,rela>vamentebarato,conpocaoningunavariacióndiurna…

•  Sedetectanconcentracionesdeestassustanciasenelcondensadodeaireexpirado, orina, sangre o plasma, líquido de lavado broncoalveolar oaspiradotraqueal...

•  Elaumentoodescensodenivelesde losmarcadores individualespuedenindicarlesiónoac>vacióndel>poespecíficodecélulaspulmonaresy,porlo tanto, pueden ser ú>les para el diagnós>co y la predicción de lamortalidad,oparamonitorizarlarespuestaaltratamiento...

CrossLJ,Ma^hayMA.Biomarkersinacutelunginjury;insightsintothepathogenesisofacutelunginjury.CritCareClin2011;27:355-377.

but a marker that is easily measured, and when combinedwith other markers is predictive of disease and/or pooroutcomes. Currently, biomarkers for ALI/ARDS remain inthe research domain; in the future, biomarkers will likelymove beyond enriching our understanding of pathogenesisinto the bedside care of the patient.

This review aims to summarize our existing knowledge ofbiomarkers and their application to clinical practice, includ-ing methods on the forefront of biomarker discovery and theapplication of current biomarkers to designing future clinicaltrials and epidemiological studies of ALI/ARDS.

Pathogenesis

Several biomarkers can be considered in the context of theirpotential origin during ALI, either the microvascular endo-thelial or alveolar epithelial barriers. Injury to and disruptionof both barriers are pathophysiological hallmarks of ALI, andeach barrier can release biomarkers that can be sampled viathe vasculature or alveoli and potentially used to assess risk,diagnose, and predict outcomes for patients with lung injury.In addition, injury to these barriers can be mediated byinflammation and activation of coagulation, processes thatcan also be detected by the measurement of biomarkers(►Table 1).1,15

Pulmonary Endothelium

Injury and activation of the pulmonary endothelium havelong been recognized as the inciting events in the pathogen-esis of ALI.15,16 Endothelial injury increases vascular perme-ability and results in accumulation of protein-rich pulmonaryedema fluid in the alveolus.17 Endothelial injury is thought tobe mediated predominantly by neutrophils that accumulatein the microvasculature, become activated by the underlyingcause of ALI (sepsis, trauma, etc.), and subsequently releasemediators capable of disrupting this barrier.15 Various medi-ators of endothelial activation, injury, and permeability can

be measured and, therefore, potentially used to diagnose andprognosticate in patients with ALI.

Angiopoietin-2Angiopoietin-2 (Ang-2) is found in Weibel-Palade bodies ofendothelial cells and is released in response to endothelialactivation by proinflammatory cytokines or local, environ-mental factors such as hypoxia.18 Release of Ang-2 andsubsequent binding to the tyrosine kinase, Tie 2 receptoron the endothelial cell renders the endothelium more re-sponsive to exogenous stimuli,18 disrupts the endothelialbarrier,19 and increases endothelial adhesion of neutro-phils.20 These observations suggest that Ang-2 could play acentral role in the pathogenesis of endothelial activation andinjury in ALI, a hypothesis that has been confirmed inexperimental studies.21

Ang-2 levels in the plasma are significantly higher inpatients with sepsis and ALI/ARDS compared with thosewith sepsis but no ALI/ARDS. Ang-2 levels correlate wellwith the pulmonary leak index and are predictive for thedevelopment of ARDS in patients both with and withoutsepsis.22 In patients admitted to a surgical intensive careunit who had not yet developed ALI/ARDS, circulating Ang-2levels were significantly higher in those who went on todevelop ALI/ARDS and also in those who did not survive.23

Finally, in patients who have ALI, Ang-2 levels have beenshown to decrease to a greater degree in response to aconservative fluid management strategy compared with pa-tients managed with a liberal fluid strategy.24

SelectinsSelectins are adhesion molecules found on the cell surface ofthe endothelium (E), leukocytes (L), and platelets (P). Thesethree types of selectins function in rolling and adhesion ofneutrophils during inflammation.25 Endothelial E-selectinexpression is upregulated in response to common risk factorsfor ALI such as sepsis and bacteremia,26 suggesting thatselectins may be potential biomarkers to predict both the

Table 1 Studied biomarkers in ALI/ARDS and their pathogenetic source or site of action

Endothelium Epithelium Alveolar-capillarymembrane

Inflammation Coagulation Metabolic

Angiopoietin-2 Surfactant proteins EF/PL total protein ratio IL-2r PAI-1 Glutathione

Selectins RAGE IL-6 Thrombomodulin Adenosine

vWf antigen Clara cell protein IL-8 Protein C Phosphatidylserine

ICAM-1 Peptidase inhibitor 3 IL-10 Tissue factor Sphingomyelin

VEGF IL-1ra Myoinositol

IGFBP-3 TNF-α

Apolipoprotein A1 IL-1β

S100 HMGB1

PPFIA1 gene

Abbreviations: HMGB1, high-mobility group box 1 protein; ICAM-1, intercellular adhesion molecule-1; IGFBP-3, insulin-like growth factor bindingprotein-3; IL, interleukin; PAI-1, plasminogen activator inhibitor-1; TNF, tumor necrosis factor; VEGF, vascular endothelial growth factor.

Seminars in Respiratory and Critical Care Medicine Vol. 34 No. 4/2013

Biomarkers of ALI/ARDS Janz, Ware538

This

doc

umen

t was

dow

nloa

ded

for p

erso

nal u

se o

nly.

Una

utho

rized

dis

tribu

tion

is s

trict

ly p

rohi

bite

d.

SeminRespirCritCareMed2013;34:537–548

Figure 1. Time course in Acute Lung Injury. Early in the course the alveoli are filled with protein rich permeability pulmonary edema. By day five to seven, there is proliferation of type II alveolar epithelial cells, leading to repithelialization and restoration of the alveolar structure or progressive fibrosis and irreversible hypoxic respiratory failure. (Redrawn from Katzenstein AA, Askin FB. Surgical Pathology of Non-neoplastic Lung)

Bhargava and Wendt Page 21

Transl Res. Author manuscript; available in PMC 2015 August 16.

Author Manuscript

Author Manuscript

Author Manuscript

Author Manuscript

ORIGINAL ARTICLE

Identification and validation of distinct biologicalphenotypes in patients with acute respiratorydistress syndrome by cluster analysisL D Bos,1,2,3 L R Schouten,1,3 L A van Vught,4 M AWiewel,4 D S Y Ong,5,6 O Cremer,6

A Artigas,7 I Martin-Loeches,8 A J Hoogendijk,4 T van der Poll,4 J Horn,1,3

N Juffermans,1,3 C S Calfee,9 M J Schultz,1,3 On behalf of the MARS consortium

▸ Additional material ispublished online only. To viewplease visit the journal online(http://dx.doi.org/10.1136/thoraxjnl-2016-209719).

For numbered affiliations seeend of article.

Correspondence toDr Lieuwe Bos, Department ofIntensive Care, AcademicMedical Center, Meibergdreef9, Amsterdam 1105 AZ, TheNetherlands; [email protected]

Received 12 November 2016Revised 27 March 2017Accepted 28 March 2017

To cite: Bos LD,Schouten LR, van Vught LA,et al. Thorax PublishedOnline First: [please includeDay Month Year]doi:10.1136/thoraxjnl-2016-209719

ABSTRACTRationale We hypothesised that patients with acuterespiratory distress syndrome (ARDS) can be clusteredbased on concentrations of plasma biomarkers and thatthe thereby identified biological phenotypes areassociated with mortality.Methods Consecutive patients with ARDS wereincluded in this prospective observational cohort study.Cluster analysis of 20 biomarkers of inflammation,coagulation and endothelial activation provided thephenotypes in a training cohort, not taking any outcomedata into account. Logistic regression with backwardselection was used to select the most predictivebiomarkers, and these predicted phenotypes werevalidated in a separate cohort. Multivariable logisticregression was used to quantify the independentassociation with mortality.Results Two phenotypes were identified in 454patients, which we named ‘uninflamed’ (N=218) and‘reactive’ (N=236). A selection of four biomarkers(interleukin-6, interferon gamma, angiopoietin 1/2 andplasminogen activator inhibitor-1) could be used toaccurately predict the phenotype in the training cohort(area under the receiver operating characteristics curve:0.98, 95% CI 0.97 to 0.99). Mortality rates were15.6% and 36.4% (p<0.001) in the training cohort and13.6% and 37.5% (p<0.001) in the validation cohort(N=207). The ‘reactive phenotype’ was independentfrom confounders associated with intensive care unitmortality (training cohort: OR 1.13, 95% CI 1.04 to1.23; validation cohort: OR 1.18, 95% CI 1.06 to 1.31).Conclusions Patients with ARDS can be clustered intotwo biological phenotypes, with different mortality rates.Four biomarkers can be used to predict the phenotypewith high accuracy. The phenotypes were very similar tothose found in cohorts derived from randomisedcontrolled trials, and these results may improve patientselection for future clinical trials targeting host responsein patients with ARDS.

INTRODUCTIONThe acute respiratory distress syndrome (ARDS) isa major complication in critically ill patients, withhigh morbidity and mortality.1–4 Despite promisingresults in preclinical experiments testing immuno-modulatory interventions in animals with lunginjury,5 6 results from clinical trials in patients with

ARDS have been disappointing so far.7–9 Differencesbetween pathological manifestations of lung injuryin animals and ARDS in patients can only partlyexplain the discrepancies between animal studiesand clinical trials.10 Furthermore, preclinical experi-ments have always used inbred animals in an effortto limit heterogeneity. Clinical trials, however, hadto rely on clinical, radiological and physiologicalparameters to diagnose and stratify ARDS.11 Thus,patients with ARDS included in clinical trials are bydefinition more heterogeneous.12

Biological subtyping of patients could improvepatient selection for clinical trials with targetedtherapies, including immunomodulatory interven-tions, as has been shown in other pulmonary andnon-pulmonary diseases.13 14 Phenotyping ofpatients with ARDS can be done using clinicalcharacteristics, causes of lung injury,15 individual orsets of biomarkers,16 or a combination of clinicaland biological variables.17 Stratification on bio-logical responses (ie, the biological phenotype) mayallow for a better selection of patients, forexample, with regard to potential benefit from acertain intervention (predictive enrichment), allow-ing exclusion of patients that have a low chance ofbenefit who may even may be harmed.18–20 Indeed,

Key messages

What is the key question?▸ Can cluster analysis of biological markers in

plasma of patients with acute respiratorydistress syndrome (ARDS) be used to identifyphenotypes with different mortality rates?

What is the bottom line?▸ There are at least two phenotypes of ARDS and

the ‘reactive’ phenotype is associated withmortality independent of severity of illness.

Why read on?▸ Pharmacological interventions in patients with

ARDS have all failed so far; this study shedslight on two phenotypes of ARDS that may betargeted differently in future randomisedcontrolled trials.

Bos LD, et al. Thorax 2017;0:1–8. doi:10.1136/thoraxjnl-2016-209719 1

Critical care Thorax Online First, published on April 27, 2017 as 10.1136/thoraxjnl-2016-209719

Copyright Article author (or their employer) 2017. Produced by BMJ Publishing Group Ltd (& BTS) under licence.

group.bmj.com on April 28, 2017 - Published by http://thorax.bmj.com/Downloaded from

scores, more organ failure and more frequently had an indirectcause for ARDS (table 2). A ‘reactive’ phenotype remained inde-pendently associated with ICU mortality after correction forAPACHE IV (OR 1.13, 95% CI 1.04 to 1.23). The addition ofother potential confounders (APPS, PaO2/FiO2 ratio, pulmonarycause for ARDS) did not change this association (OR remained1.11). The difference in mortality between the biological pheno-types was also independent of the Berlin classification of ARDS(figure 2, OR 3.1, 95% CI 2.0 to 4.8).

Prediction of phenotypes based on a limited set ofbiomarkersThe plasma concentration of IL-6, IFN-γ, ANG1/2 and PAI-1could be used to accurately discriminate between the two bio-logical phenotypes in the training cohort (figure 3; area underthe receiver operating characteristics curve: 0.98, 95% CI 0.97to 0.99). The regression coefficients can be found in the onlinesupplementary table S3. Prediction of the phenotype by rou-tinely available variables that were significantly differentbetween the phenotypes (APACHE IV, age, lactate, albumin,bicarbonate, mean arterial pressure, bicarbonate, platelets, Creactive protein, maximum inspiratory pressure, positive endexpiratory pressure (PEEP) and PaO2/FiO2) had a significantlylower accuracy than that of the biomarkers (figure 3; p<0.001).The same discrimination could also be obtained by only usingplasma albumin and bicarbonate concentration.

Association of phenotypes with clinical outcome in thevalidation cohortIn the validation cohort, the predicted ‘uninflamed’ and ‘react-ive’ phenotype had a mortality rate of 13.6% and 37.5%(p<0.001), respectively. The differences in clinical character-istics were comparable to those found in the training cohort

Figure 1 Heatmap of phenotypes. Columns: biomarkers. Rows:patients. First column: green blocks: ‘uninflamed phenotype’; red:‘reactive phenotype’. Second column: patients that died are indicatedwith black, surviving patients with grey. Heat map: a higherconcentration, in comparison to the other included patients is indicatedwith red, while a lower concentration is indicated by blue. ANG-1,angiopoietin-1; ICAM-1, intercellular adhesion molecule-1; ICU,intensive care unit; IFN, interferon; IL, interleukin; MMP8, matrixmetalloproteinase-8; PAI-1, plasminogen activator inhibitor 1; TIMP1,tissue inhibitor of metalloprotease 1; TNF, tumour necrosis factor; tPA,tissue plasminogen activator.

Table 2 Phenotypes versus clinical characteristics in trainingcohort

UninflamedphenotypeN=218

ReactivephenotypeN=236 p Value

Age 62 (53.2–72) 60 (49–70) 0.037Male 137 (62.8) 151 (64) 0.85APACHE IV score 69 (58–91) 93 (74–113) <0.001APACHE IV acutephysiology score

57 (45–75) 80 (65–105) <0.001

Admission typeMedical 143 (65.6) 167 (70.8) 0.002Elective surgery 46 (21.1) 21 (8.9)Emergency surgery 29 (13.3) 48 (20.3)

Chronic renal insufficiency 17 (7.8) 31 (13.1) 0.08Chronic respiratoryinsufficiency

24 (11) 8 (3.4) 0.002

COPD 26 (11.9) 18 (7.6) 0.16Diabetes mellitus 33 (15.1) 33 (14) 0.80Immune deficiency 37 (17) 49 (20.8) 0.34Current drinking status(alcohol)

19 (8.7) 29 (12.3) 0.23

Systemic corticosteroids(before ICU)

32 (14.7) 24 (10.2) 0.15

Direct hit for ARDS 145 (66.5) 134 (56.8) 0.04Berlin classificationMild 85 (39) 68 (28.8) 0.07Moderate 103 (47.2) 128 (54.2)Severe 30 (13.8) 40 (16.9)

Maximal inspiratorypressure

20 (16–26) 26 (21–33) <0.001

PaO2/FiO2 177.8 (136–256) 178 (133–223) 0.18PEEP 8 (5–11) 10 (8–14) <0.001Tidal volume/kg predictedbodyweight

7.1 (6.3–8.1) 7.1 (6.2–8.3) 0.92

APPS 5 (4–6) 5 (5–7) 0.008SOFA: circulation 3 (1–4) 4 (3–4) <0.001SOFA: CNS 0 (0–1) 0 (0–1) 0.97SOFA: coagulation 0 (0–1) 1 (0–2) <0.001SOFA: liver 0 (0–0) 0 (0–1) <0.001SOFA: renal 0 (0–1) 1 (0–3) <0.001SOFA: respiratory 3 (3–4) 3 (3–4) 0.05SOFA: total score 7 (5–9) 10 (8–12) <0.001Days on mechanicalventilation

6 (3–10) 7 (4–14.5) 0.004

ICU length of stay 7 (4–12) 10 (5–19) 0.006Days free of MV at day 28 21 (11–25) 9 (0–21) <0.001ICU mortality 34 (15.6) 86 (36.4) <0.00130 day mortality 47 (21.6) 89 (37.7) <0.001

Data are presented as the median with IQR for continuous variables and as numberwith percentage for categorical variables. The p value is calculated by theKruskal-Wallis test for continuous variables and by Fisher’s exact for categoricalvariables. Definitions for the variables are given in the definition table at the end ofthe paper.APACHE, Acute Physiology and Chronic Health Evaluation; APPS, Age, PaO2/FiO2 andPlateau pressure Score; ARDS, acute respiratory distress syndrome; CNS, centralnervous system; ICU, intensive care unit; MV, mechanical ventilation; PEEP, positiveend expiratory pressure; SOFA, sepsis-related organ failure assessment.

4 Bos LD, et al. Thorax 2017;0:1–8. doi:10.1136/thoraxjnl-2016-209719

Critical care


BosLD,etal.Thorax2017;0:1–8.doi:10.1136/thoraxjnl-2016-209719

“EnriquecimientodelPronós'co”

“EnriquecimientoPredic'vo”

“MedicinaEstra'ficada”ó“MedicinadePrecisión”

IncrementarbeneficiosyReducirriesgos

(online supplementary table S4, figures S3 and S4). A ‘reactive’phenotype was independently associated with ICU mortalityafter correction for APACHE IV (OR 1.18, 95% CI 1.06 to1.31) and for the Berlin classification for severity of ARDS (OR3.8, 95% CI 2.0 to 7.2).

Confounding factors and association of phenotype withhospital mortalityPatients in whom the sample was taken the day before or afterthe diagnosis of ARDS were not more or less likely to be classi-fied as having an ‘uninflamed’ or ‘reactive’ phenotype (p=0.34for sample taken before and p=0.13 for taken after). A sensitiv-ity analysis for patients with a pulmonary cause for ARDS aloneand for patients without chronic respiratory failure also showedno change in OR. Sensitivity analysis on the association betweenthe phenotypes and hospital mortality showed similar results asin the primary analysis (OR 1.10, 95% CI 1.03 to 1.18).Exposure to steroids on the ICU (147/700, 21%) was associatedwith a higher likelihood of a ‘reactive’ phenotype (OR 2.1, 95%CI 1.5 to 3.1).

DISCUSSIONTwo biologically distinct clusters of patients with ARDS couldbe identified. Outcome data were not taken into account whenseparating these clusters. Patients with the ‘reactive’ phenotypewere approximately twice as likely to die during their stay in theICU. Importantly, the biological phenotypes contained add-itional information compared with two mortality predictionscores and the Berlin classification for ARDS, and these resultswere validated in an independent group of patients. A ‘reactive’

phenotype could be predicted with the plasma concentration offour biomarkers and routinely available variables led to a lessaccurate prediction. We speculate that these biological pheno-types might be used to include patients for the appropriatepharmacological therapy in clinical trials.

To our knowledge, this is the first study to cluster patientswith ARDS based on biomarker concentrations alone. Clusteranalysis maximises the differences between patients, withouttaking the clinical outcome of a patient into account. Therefore,it is very different from, for example, logistic regression withsingle biomarkers.34 Calfee et al previously showed two distinctclusters of patients with ARDS within the cohorts of two largeclinical trials.17 Both clinical and biomarker data were used tocluster the patients, and the clusters responded differently torandomly allocated changes in ventilator settings. Interestingly,mortality in the phenotypes that we identified was similar tothat found in the phenotypes in their study (eg, ±20% vs±45%). The phenotypes found in the Calfee study were repli-cated in another RCT population, in which the influence offluid resuscitation management was tested.21 That study alsorevealed similar mortality rates and found that the response torandomly allocated fluid management differed per phenotype.All three studies found an increase in plasma IL-8 and PAI-1 con-centration and a decrease in bicarbonate concentration in the‘reactive’ or ‘hyperinflammatory’ phenotype. Therefore, we canspeculate that the identified phenotypes could be the samebetween this observational study and the three RCTs, even thoughthe prevalence of a ‘reactive’ phenotype is higher in our study.This finding would be notable because of the differences betweenthe studies; observation and interventional, recruitment on

Figure 2 Orthogonality ofphenotypes and Berlin classification.Intensive care unit (ICU) mortality perphenotype and Berlin classification.Boxes indicate phenotypes and thetraining or validation cohort, separatebars Berlin categories. Differences inmortality between the ‘reactive’phenotype and ‘uninflamed’ phenotypewere independent of the Berlinclassification of acute respiratorydistress syndrome (OR 3.1, 95% CI 2.0to 4.8) in the training cohort and inthe validation cohort (OR 3.8, 95% CI2.0 to 7.2).


Critical care


(online supplementary table S4, figures S3 and S4). A ‘reactive’phenotype was independently associated with ICU mortalityafter correction for APACHE IV (OR 1.18, 95% CI 1.06 to1.31) and for the Berlin classification for severity of ARDS (OR3.8, 95% CI 2.0 to 7.2).

Confounding factors and association of phenotype withhospital mortalityPatients in whom the sample was taken the day before or afterthe diagnosis of ARDS were not more or less likely to be classi-fied as having an ‘uninflamed’ or ‘reactive’ phenotype (p=0.34for sample taken before and p=0.13 for taken after). A sensitiv-ity analysis for patients with a pulmonary cause for ARDS aloneand for patients without chronic respiratory failure also showedno change in OR. Sensitivity analysis on the association betweenthe phenotypes and hospital mortality showed similar results asin the primary analysis (OR 1.10, 95% CI 1.03 to 1.18).Exposure to steroids on the ICU (147/700, 21%) was associatedwith a higher likelihood of a ‘reactive’ phenotype (OR 2.1, 95%CI 1.5 to 3.1).

DISCUSSIONTwo biologically distinct clusters of patients with ARDS couldbe identified. Outcome data were not taken into account whenseparating these clusters. Patients with the ‘reactive’ phenotypewere approximately twice as likely to die during their stay in theICU. Importantly, the biological phenotypes contained add-itional information compared with two mortality predictionscores and the Berlin classification for ARDS, and these resultswere validated in an independent group of patients. A ‘reactive’

phenotype could be predicted with the plasma concentration offour biomarkers and routinely available variables led to a lessaccurate prediction. We speculate that these biological pheno-types might be used to include patients for the appropriatepharmacological therapy in clinical trials.

To our knowledge, this is the first study to cluster patientswith ARDS based on biomarker concentrations alone. Clusteranalysis maximises the differences between patients, withouttaking the clinical outcome of a patient into account. Therefore,it is very different from, for example, logistic regression withsingle biomarkers.34 Calfee et al previously showed two distinctclusters of patients with ARDS within the cohorts of two largeclinical trials.17 Both clinical and biomarker data were used tocluster the patients, and the clusters responded differently torandomly allocated changes in ventilator settings. Interestingly,mortality in the phenotypes that we identified was similar tothat found in the phenotypes in their study (eg, ±20% vs±45%). The phenotypes found in the Calfee study were repli-cated in another RCT population, in which the influence offluid resuscitation management was tested.21 That study alsorevealed similar mortality rates and found that the response torandomly allocated fluid management differed per phenotype.All three studies found an increase in plasma IL-8 and PAI-1 con-centration and a decrease in bicarbonate concentration in the‘reactive’ or ‘hyperinflammatory’ phenotype. Therefore, we canspeculate that the identified phenotypes could be the samebetween this observational study and the three RCTs, even thoughthe prevalence of a ‘reactive’ phenotype is higher in our study.This finding would be notable because of the differences betweenthe studies; observation and interventional, recruitment on

Figure 2 Orthogonality ofphenotypes and Berlin classification.Intensive care unit (ICU) mortality perphenotype and Berlin classification.Boxes indicate phenotypes and thetraining or validation cohort, separatebars Berlin categories. Differences inmortality between the ‘reactive’phenotype and ‘uninflamed’ phenotypewere independent of the Berlinclassification of acute respiratorydistress syndrome (OR 3.1, 95% CI 2.0to 4.8) in the training cohort and inthe validation cohort (OR 3.8, 95% CI2.0 to 7.2).


Critical care



different continents and inclusion periods spanning >15 yearsbetween the ARMA trial and this study. Furthermore, the Calfeeet al and Famous et al studies used both clinical and biomarkersdata while we limited the analysis to biomarker data alone. Finally,the studies used different methods of clustering. The fact that theresults show the identification of very similar phenotypes suggeststhe underlying identified biological signal is very strong.

The ‘reactive’ phenotype had a higher ICU mortality andmight be used to select more severely ill patients for clinicaltrials (prognostic enrichment). Importantly, this association wasindependent of APACHE IV score, a frequently used validatedand repeatedly calibrated prognostic score for mortality in theICU. There was also added value of the biological clusters to theBerlin definition in the prediction of mortality. The resultsimply that the ‘reactive’ phenotype is not just a more severeform of ARDS and most definitely not captures the same gradesof severity as the Berlin classification. This finding also illustratesthat phenotypic clustering provides additional information, ontop of the more frequently used clinical and physiological infor-mation. The PaO2/FiO2 is, until now, the only characteristic thathas been used for phenotypic inclusion into clinical trials withpatients with ARDS,24 25 35 with moderate success. Severalinterventions had positive effects on mortality in a selectedgroup of ARDS with a low PaO2/FiO2 ratio.24 25 35

Interestingly, these were all interventions that aimed for physio-logical changes to improve oxygenation. In stark contrast,pharmacological interventions aimed at the immune system haverepeatedly showed no benefit when applied to unselected ARDSpatient groups.7–9

We speculate that the phenotypes, derived from biologicaldata alone, as identified in this study could be used to targetpharmacological interventions to those patients that benefitmost in future clinical trials. Improved patient selection and tar-geted intervention is the premise of this so-called ‘stratifiedmedicine’ or ‘precision medicine’.36 The efficacy of some phar-macotherapies could potentially be improved by correctly select-ing the subgroups of patients that show molecular signs ofsusceptibility (predictive enrichment). Simultaneously, thisapproach may limit exposure of patients that would not benefit,but would have side effects. Thus, stratified medicine mayincrease efficiency of a drug in two ways: increase benefit anddecrease harm. As for the phenotypes identified in this study,we postulate that the ‘reactive’ phenotype might benefit mostfrom immunomodulatory interventions, such as corticosteroids,macrolides or others. On the other hand, the ‘uninflamed’phenotype may be less likely to benefit from these approachesand/or may be more likely to be harmed, as mortality is infre-quent and there is little inflammatory response.

ARDS phenotyping could allow for a more targeted pharma-cological intervention in clinical trials and, if shown to be bene-ficial, in clinical practice. There are, however, severalprerequisites before that hypothesis can be tested. The first, pre-diction of cluster membership by a minimal number of biomar-kers, is explored in this paper. IL-6, IFN-γ, ANG1/2 and PAI-1concentrations in plasma drawn at the moment of ARDS diag-nosis were sufficient to discriminate between patients with andwithout a ‘reactive’ phenotype. A four-biomarker assay is suffi-ciently small to allow for phenotyping of patients in clinical

Figure 3 Discrimination of biologicalphenotype based on a limited set ofbiomarkers in the training cohort.Receiver operating characteristics curvefor the biological phenotype based on(1) biomarkers depicted in black:plasma concentrations of interleukin-6,interferon-γ, angiopoietin-1/2 andplasminogen activator inhibitor-1 (seeonline supplementary table S3) and (2)routinely available clinical variablesdepicted in grey; the same accuracycould be obtained with albumin andbicarbonate only. AUC, area under thecurve.


Critical care


different continents and inclusion periods spanning >15 yearsbetween the ARMA trial and this study. Furthermore, the Calfeeet al and Famous et al studies used both clinical and biomarkersdata while we limited the analysis to biomarker data alone. Finally,the studies used different methods of clustering. The fact that theresults show the identification of very similar phenotypes suggeststhe underlying identified biological signal is very strong.

The ‘reactive’ phenotype had a higher ICU mortality andmight be used to select more severely ill patients for clinicaltrials (prognostic enrichment). Importantly, this association wasindependent of APACHE IV score, a frequently used validatedand repeatedly calibrated prognostic score for mortality in theICU. There was also added value of the biological clusters to theBerlin definition in the prediction of mortality. The resultsimply that the ‘reactive’ phenotype is not just a more severeform of ARDS and most definitely not captures the same gradesof severity as the Berlin classification. This finding also illustratesthat phenotypic clustering provides additional information, ontop of the more frequently used clinical and physiological infor-mation. The PaO2/FiO2 is, until now, the only characteristic thathas been used for phenotypic inclusion into clinical trials withpatients with ARDS,24 25 35 with moderate success. Severalinterventions had positive effects on mortality in a selectedgroup of ARDS with a low PaO2/FiO2 ratio.24 25 35

Interestingly, these were all interventions that aimed for physio-logical changes to improve oxygenation. In stark contrast,pharmacological interventions aimed at the immune system haverepeatedly showed no benefit when applied to unselected ARDSpatient groups.7–9

We speculate that the phenotypes, derived from biologicaldata alone, as identified in this study could be used to targetpharmacological interventions to those patients that benefitmost in future clinical trials. Improved patient selection and tar-geted intervention is the premise of this so-called ‘stratifiedmedicine’ or ‘precision medicine’.36 The efficacy of some phar-macotherapies could potentially be improved by correctly select-ing the subgroups of patients that show molecular signs ofsusceptibility (predictive enrichment). Simultaneously, thisapproach may limit exposure of patients that would not benefit,but would have side effects. Thus, stratified medicine mayincrease efficiency of a drug in two ways: increase benefit anddecrease harm. As for the phenotypes identified in this study,we postulate that the ‘reactive’ phenotype might benefit mostfrom immunomodulatory interventions, such as corticosteroids,macrolides or others. On the other hand, the ‘uninflamed’phenotype may be less likely to benefit from these approachesand/or may be more likely to be harmed, as mortality is infre-quent and there is little inflammatory response.

ARDS phenotyping could allow for a more targeted pharma-cological intervention in clinical trials and, if shown to be bene-ficial, in clinical practice. There are, however, severalprerequisites before that hypothesis can be tested. The first, pre-diction of cluster membership by a minimal number of biomar-kers, is explored in this paper. IL-6, IFN-γ, ANG1/2 and PAI-1concentrations in plasma drawn at the moment of ARDS diag-nosis were sufficient to discriminate between patients with andwithout a ‘reactive’ phenotype. A four-biomarker assay is suffi-ciently small to allow for phenotyping of patients in clinical

Figure 3 Discrimination of biologicalphenotype based on a limited set ofbiomarkers in the training cohort.Receiver operating characteristics curvefor the biological phenotype based on(1) biomarkers depicted in black:plasma concentrations of interleukin-6,interferon-γ, angiopoietin-1/2 andplasminogen activator inhibitor-1 (seeonline supplementary table S3) and (2)routinely available clinical variablesdepicted in grey; the same accuracycould be obtained with albumin andbicarbonate only. AUC, area under thecurve.


Critical care



In 2010, Ware and colleagues measured the EF/PL ratio in390 mechanically ventilated patients with acute pulmonaryedema diagnosed as either acute lung injury, cardiogenic pulmo-nary edema or mixed etiology based on expert consensus at theend of hospitalization. EF/PL had an area under the curve(AUC) of 0.84 for differentiating between ARDS and cardio-genic pulmonary edema [23]. Using a cutoff of ‡0.65 yielded asensitivity of 81% and specificity of 81% for diagnosing ARDS.Moreover, patients with EF/PL ‡0.65 had significantly increasedin-hospital mortality, suggesting that EF/PL may be a useful bio-marker for both diagnosis and prognostication in ARDS.

Vascular endothelial biomarkersVascular endothelial injury is caused by activation of inflamma-tory and coagulation cascades. Activation of endothelial cells bycirculating mediators leads to increased expression of cell sur-face molecules important for leukocyte adhesion, contributingto leukocyte accumulation and transmigration [24]. Activatedpolymorphonuclear lymphocytes that collect in microvesselscan also release mediators that increase vascular permeability.Along with these leukocyte signals, inflammatory mediatorssuch as TNF, thrombin and VEGF disrupt endothelial-cadherin bonds and contribute to the vascular leak underlying

edema formation in ARDS. Platelets are also believed to con-tribute to endothelial injury through release of cytokines andby leading to fibrin clots.

Angiopoietin 2The angiopoietin signaling pathway is, along with VEGF, one ofthe two signaling pathways specific to endothelial cells. Ang-1 isan angiogenic factor that interacts with the endothelial tyrosinekinase receptor Tie-2 to stabilize endothelial cells during inflam-mation, injury and angiogenesis. Ang-1 helps limit endothelialcell permeability, inhibits expression of tissue factor (TF) andsuppresses leukocyte adhesion [25]. Ang-2 is a naturally occurringantagonist to Ang-1 released by activated endothelial cells inresponse to inflammation and injury. Ang-2 increases endothelialjunction instability, sensitizes endothelial cells to inflammatorystimuli and aids in inflammatory remodeling of the vasculature.

In a prospective study of 230 patients admitted to the inten-sive care unit without initial evidence of ARDS, baselineplasma levels of Ang-2 and the ratio of Ang-2 to Ang-1 weresignificantly associated with the development of ARDS [26].Adding Ang-2 to the Lung Injury Prediction Score, a validatedprediction tool for the development of ARDS [27], improvedthe score’s predictive value.

Endothelial injury

Epithelial injury

Activated inflammatorycascade

Coagulation andfibrinolysis

Alveolar-capillaryBarrier permeability

Fibrosis

Ang-2, VWF, VEGF, ICAM-1, selectins

Surfactant proteins, RAGE, CC16, KL-6,ICAM-1

HMGB1, IL-1β, IL-6, IL-8, IL-10, TNF-α,soluble TNF-α receptors

PAI-I, protein C, thrombomodulin

EF/PL protein ratio

PCPIII

ApoptosisFas/FasL

MacrophageMonocyte

Fibroblast

Red blood cell

Fibrin

Neutrophil

Alveolus

Type I cell

Club cell

Type II cell

Protein-richedema fluid

Capillary

Figure 1. Schematic of an injured alveolus in the acute phase of acute respiratory distress syndrome. Candidate biomarkers areorganized by the cellular injury pathways central to lung injury pathogenesis.CC16: Club cell protein 16; EF/PL: Edema fluid to plasma protein ratio; HMGB1: High mobility group box protein 1; KL-6: Krebs von denLungen-6; PAI-1: Plasminogen activator inhibitor-1; PCPIII: Procollagen III peptide; RAGE: Receptor for advanced glycation end-products;VWF: Von Willebrand factor.

Biomarkers in ARDS Review

informahealthcare.com 575

Expe

rt R

evie

w o

f Res

pira

tory

Med

icin

e D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y U

nive

rsity

of Q

ueen

slan

d on

03/

14/1

5Fo

r per

sona

l use

onl

y.

WalterJM,WilsonJ,WareLB.Biomarkersinacuterespiratorydistresssyndrome:frompathobiologytoimprovingpa-entcare.ExpertRevRespirMed2014;8(5):573-586.

BhargavaM,WendtCH.Biomarkersinacutelunginjury.TranslRes2012;159:205-217.

Células epiteliales tipo I •  La fase exuda>va se caracteriza por la formación de edemapulmonar debido al dañoalveolardifuso…

•  La lesiónde lascélulasalveolaresde>po Iasí comodisrrupciónde labarreraepitelialpromueveelcúmulodelíquidoenelespacioalveolareinters>cio...

•  En la superficie basal de las células de >po I, se ha iden>ficado el receptor paraproductos finales de glicación avanzada (RAGE: receptor for advanced glyca'on endproducts),responsabledelapropagacióndelarespuestainflamatoriaatravésdeNF-kB,aumentando así la producción de citoquinas proinflamatorias, especies reac>vas deoxígenoyproteasas...

•  Los niveles plasmá>cos de RAGE incrementan en el día 1 del SIRA, el valor esproporcionalalamagnitudoseveridaddelSIRA,noan>cipadesenlace...

JabaudonM,Fur>erE,RoszykL,ChalusE,etal.Solubleformofthereceptorforadvancedglyca-onendproductsisamarkerofacutelunginjurybutnotofseveresepsisincri-callyillpa-ents.CritCareMed2011;39:480-488.

Células epiteliales tipo II •  Proteínasespecíficasdelsurfactante(SP)

•  SP-A•  SP-D

•  SP-D•  Granriesgodemuerte•  Menosdíaslibresdeven>laciónmecánica•  Menosdíaslibresdedisfunciónorgánica•  Elevacióntempranapeordesenlaceclínico

• ⬇SP-D(líquidopulmonar)/⬆SP-Asérico•  Enfermedadgrave•  Peordesenlaceclínico

EisnerMD,ParsonsP,Ma^hayMA,WareL,etal.Plasmasurfactantproteinlevelsandclinicaloutcomesinpa-entswithacutelunginjury.Thorax2003;58:983-988.

• Clubcellsecretoryprotein(CCSP)•  Regulacióndelarespuestainflamatoria• ⬆valoresséricosasociadaapeordesenlace

•  sICAM-1(solubleintracellularadhesionmolecule-1)•  Marcadordedañodecélulasapitelialesyendotelio•  Incrementoprogresivoasociadoapobredesenlaceclínico•  Incrementodentrodelosprimerostresdías…altoriesgodemuerte

DetermannRM,MilloJL,WaddyS,Lu^erR,etal.PlasmaCC16levelsareassociatedwithdevelopmentofALI/ARDSinpa-entswithven-lator-associatedpneumonia:aretrospec-veobserva-onalstudy.BMCPulm2009;9:49.

CalfeeCS,EisnerMD,ParsonsPE,ThompsonBT,etal.Solubleintercellularadhesionmolecule-1andclinicaloutcomesinpa-entswithacutelunginjury.IntensiveCareMed2009;35:248-257.

RESEARCH Open Access

Biomarkers of lung epithelial injury andinflammation distinguish severe sepsis patientswith acute respiratory distress syndromeLorraine B Ware1,2*, Tatsuki Koyama3, Zhiguo Zhao3, David R Janz1, Nancy Wickersham1, Gordon R Bernard1,Addison K May4, Carolyn S Calfee5 and Michael A Matthay5

Abstract

Introduction: Despite recent modifications, the clinical definition of the acute respiratory distress syndrome (ARDS)remains non-specific, leading to under-diagnosis and under-treatment. This study was designed to test the hypothesisthat a biomarker panel would be useful for biologic confirmation of the clinical diagnosis of ARDS in patients at risk ofdeveloping ARDS due to severe sepsis.

Methods: This was a retrospective case control study of 100 patients with severe sepsis and no evidence of ARDScompared to 100 patients with severe sepsis and evidence of ARDS on at least two of their first four ICU days. A panelthat included 11 biomarkers of inflammation, fibroblast activation, proteolytic injury, endothelial injury, and lungepithelial injury was measured in plasma from the morning of ICU day two. A backward elimination model buildingstrategy on 1,000 bootstrapped data was used to select the best performing biomarkers for further consideration in alogistic regression model for diagnosis of ARDS.

Results: Using the five best-performing biomarkers (surfactant protein-D (SP-D), receptor for advanced glycationend-products (RAGE), interleukin-8 (IL-8), club cell secretory protein (CC-16), and interleukin-6 (IL-6)) the area underthe receiver operator characteristic curve (AUC) was 0.75 (95% CI: 0.7 to 0.84) for the diagnosis of ARDS. The AUCimproved to 0.82 (95% CI: 0.77 to 0.90) for diagnosis of severe ARDS, defined as ARDS present on all four of thefirst four ICU days.

Conclusions: Abnormal levels of five plasma biomarkers including three biomarkers generated by lung epithelium(SP-D, RAGE, CC-16) provided excellent discrimination for diagnosis of ARDS in patients with severe sepsis. Alteredlevels of plasma biomarkers may be useful biologic confirmation of the diagnosis of ARDS in patients with sepsis,and also potentially for selecting patients for clinical trials that are designed to reduce lung epithelial injury.

IntroductionThe Acute Respiratory Distress Syndrome (ARDS) is acommon clinical syndrome of acute lung inflammation,non-cardiogenic pulmonary edema and acute respira-tory failure [1]. Despite recent modifications [2] to theAmerican European Consensus Conference (AECC)definition [3], the clinical definition of ARDS remains

non-specific and is not uniformly applied. As a result,ARDS remains underdiagnosed and undertreated.The discovery and validation of biomarkers of myocar-

dial injury and ventricular overload such as troponinand brain-natriuretic peptide (BNP) has transformed thediagnosis, management and design of clinical trials inconditions such as myocardial infarction and congestiveheart failure. In a similar way, identification of plasmabiomarkers that facilitate diagnosis of ARDS could improveclinical care, enhance our understanding of pathophysi-ology, and could be used to enroll a more homogeneousgroup of patients into clinical trials of new therapies,increasing the likelihood of detecting a treatment effect.Although several plasma biomarkers have been studied in

* Correspondence: [email protected] of Allergy, Pulmonary and Critical Care Medicine, Department ofMedicine, Vanderbilt University School of Medicine, T1218 MCN, 1161 21stAvenue S, Nashville, TN 37232-2650, USA2Department of Pathology, Microbiology and Immunology, VanderbiltUniversity School of Medicine, Nashville, TN, USAFull list of author information is available at the end of the article

© 2013 Ware et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.

Ware et al. Critical Care 2013, 17:R253http://ccforum.com/content/17/5/R253

repeated the analysis using only the most severe cases(n = 66) who met ARDS criteria on all four days in theICU and their matched controls. In this analysis themodel with the top five biomarkers (SP-D, RAGE, IL-8,CC-16 and IL-6) had an AUC of 0.82 (95% CI: 0.77 to0.90) (Table 3, Figure 1) compared to single biomarkerperformance AUCs ranging from 0.63 (IL-6) to 0.72(SP-D). We also assessed the sensitivity and specificityof the diagnostic models. Setting the sensitivity at 70%,the specificity of the diagnostic model that included allpatients was 68%. Specificity improved to 75% and 83%when the analysis was restricted to patients with ARDSon enrollment day or to patients with the ARDS on allfour study days, respectively.

DiscussionThe diagnosis of ARDS is based on clinical definitionsthat lack both sensitivity and specificity. The goal of thecurrent study was to test the performance of a panel ofbiomarkers for the diagnosis of ARDS and to test if these

plasma markers would provide biologic confirmation ofthe clinical diagnosis. To reduce clinical heterogeneity, wefocused on patients with severe sepsis, the most commonand most lethal underlying etiology of ARDS [21]. Ashypothesized, a logistic regression model that utilized apanel of biomarkers had substantially superior perform-ance to single biomarkers for differentiating sepsis patientswith ARDS from those without ARDS as evaluated byROC curve analysis. From among the 11 biomarkers testedin this exploratory study, a panel that included the five top-performing biomarkers (SP-D, RAGE, IL-8, CC-16 andIL-6) had an AUC of 0.75 (95% CI: 0.7 to 0.84) for thediagnosis of ARDS. Performance of the biomarker panelwas further enhanced when only patients with ARDS atthe time of blood draw (AUC 0.78) or patients with themost severe ARDS (AUC 0.82) were considered.The best performing biomarkers for the diagnosis of

ARDS in the current study of patients with severe sepsiswere different from the best performing biomarkers identi-fied in a similar study in patients with severe trauma [6],

Table 2 Comparison of plasma biomarker levels between 100 severe sepsis patients with ARDS (cases) and 100 severesepsis patients without ARDS (controls)Biomarker Number ARDS cases number = 100 Controls number = 100 P Value

SP-D (ng/ml) 200 86 (46 to 159) 43 (28 to 77) <0.001

RAGE (pg/ml) 200 1,844 (1,060 to 3,737) 1,232 (766 to 2147) <0.001

IL-8 (pg/ml) 200 27.2 (14.2 to 119.5) 16.8 (9.3 to 45.6) 0.006

CC16 (ng/ml) 200 9.2 (5.0 to 15.8) 13.1 (6.3 to 27.8) 0.013

IL-6 (pg/ml) 200 283 (92 to 803) 142 (57 to 550) 0.023

PCPIII (ug/ml) 200 11.5 (6.9 to 18.7) 11.5 (6.7 to 34.2) 0.25

BNP (pg/ml) 200 404 (264 to 748) 412 (212 to 712) 0.38

MMP-9 (ng/ml) 195 151 (67 to 300) 163 (81 to 337) 0.39

MMP-1 (ng/ml) 195 28 (15 to 51) 31 (17 to 56) 0.44

Ang2 (ng/ml) 200 13 (7 to 21) 12 (7 to 20) 0.67

MMP-3 (ng/ml) 195 22 (15 to 35) 22 (14 to43) 0.78

Data as median (interquartile range). ARDS, acute respiratory distress syndrome; SP-D, surfactant protein D; RAGE, receptor for advanced glycation endproducts;IL-8, interleukin 8; CC16, club cell protein-16; IL-6, interleukin 6; PCPIII, procollagen peptide III; BNP, brain natriuretic peptide; MMP-9, matrix metalloprotease 9;MMP-1, matrix metalloprotease 1; Ang2, angiopoietin 2; MMP-3, matrix metalloprotease 3.

Table 3 Comparison of models for diagnosis of ARDS using single biomarkers to a combined model utilizing the topfive performing biomarkersModel All data (100 pairs) Enrollment day cases (91 pairs) Severe cases only (66 pairs)

AUC (95% CI)a AUC (95% CI)a AUC (95% CI)a

Single marker models

SPD 0.69 (0.6, 0.76) 0.71 (0.63, 0.79) 0.72 (0.62, 0.81)

RAGE 0.64 (0.56, 0.72) 0.68 (0.6, 0.75) 0.67 (0.57, 0.76)

IL8 0.61 (0.54, 0.69) 0.63 (0.55, 0.7) 0.64 (0.55, 0.73)

CC16 0.60 (0.52, 0.68) 0.60 (0.52, 0.68) 0.64 (0.55, 0.74)

IL6 0.59 (0.52, 0.67) 0.61 (0.53, 0.69) 0.63 (0.53, 0.72)

Multivariable model (includes SPD, RAGE, IL-8, CC16, IL6) 0.75 (0.7, 0.84) 0.78 (0.74, 0.87) 0.82 (0.77, 0.9)aAll AUCs are bootstrap bias-corrected; all 95% CI are bootstrap CIs. AUC, area under the receiver operator characteristic curve; CI, confidence interval; ARDS, acuterespiratory distress syndrome; SP-D, surfactant protein D; RAGE, receptor for advanced glycation endproducts; IL-8, interleukin 8; CC16, club cell protein-16; IL-6,interleukin 6.

Ware et al. Critical Care 2013, 17:R253 Page 4 of 7http://ccforum.com/content/17/5/R253

Wareetal.Cri>calCare2013,17:R253

although there was some overlap. In the trauma study, thebest performing biomarkers were RAGE, PCPIII, BNP,ANG2, IL10, TNF-α, and IL8 with an AUC of 0.86 fordifferentiating patients with ARDS from critically illtrauma patients without ARDS. Two biomarkers, RAGEand IL-8, contributed to diagnostic models in both studies.In trauma patients, biomarkers of fibroblast activation(PCPIII), endothelial injury (ANG2), other inflammatorymarkers (IL-10, TNF) and heart failure (BNP, lower inARDS) were also useful for differentiating cases from con-trols whereas in sepsis, biomarkers of lung epithelial injury(SPD, RAGE, CC16) and inflammation (IL6, IL8) predom-inate. These differences may reflect important differencesin the pathophysiology of both the underlying conditions(trauma and sepsis) as well as differences in the patho-physiology of ARDS in these different clinical settings.

Three of the top five performing biomarkers in thecurrent study were biomarkers of lung epithelial injury.SP-D is a normal constituent of surfactant that is pro-duced almost exclusively by the alveolar epithelial type IIcell. Elevated levels of SP-D have been reported in thecirculation in patients with ARDS compared to thosewith hydrostatic pulmonary edema [22] and higher levelshave been independently associated with 180-day mortal-ity and reduced ventilator and organ-failure free days inpatients with ARDS [23]. RAGE, although ubiquitouslyexpressed, is most highly expressed by the type I alveolarcell [24]. Similar to SP-D, higher plasma levels of RAGEhave been associated with adverse outcomes in patientswith ARDS [25]. CC16 is a small 16 kDa protein secretedby the club cells of the distal airway (previously known asClara cells). In contrast to other lung epithelial markers,lower levels of CC16 have previously been associated withARDS [26] although one small study reported an increasein levels at the time of onset of ARDS in patients withventilator-associated pneumonia [27]. In the current study,lower levels of CC16 were associated with the diagnosis ofARDS. Taken together with prior studies of lung epithelialmarkers and prognosis of ARDS, the current findings indi-cate that alterations in the plasma levels of biomarkers oflung epithelial injury are key features that can be used todifferentiate both the presence and the severity of ARDSin patients with sepsis. Measures of lung epithelial injurymight be particularly useful in selecting patients likely tobenefit from lung-epithelial targeted therapies such askeratinocyte growth factor [28-30] in future clinical trialsin ARDS.This study has both strengths and limitations. Major

strengths include the detailed, daily prospective patientphenotyping for ARDS as part of the VALID cohort study,and the matching of cases and controls for important clin-ical characteristics, such as age, gender, severity of illnessand number of non-pulmonary organ failures. This strin-gent matching of cases and controls eliminates severity ofillness as the primary determinant of differences in bio-marker levels in this exploratory study, thus making itmore likely that biomarkers that truly reflect the presenceof ARDS have been identified. Limitations of the study in-clude the retrospective single center design and the casecontrol study design. A case control design is more likelyto overestimate the association between a given biomarkerand the diagnosis of ARDS compared to a prospective co-hort design. In addition, the study included only 11 bio-markers and thus was not an exhaustive examination ofall potential biomarkers that might be useful for the diag-nosis of ARDS. Nevertheless, the simultaneous compari-son of performance of 11 biomarkers for the diagnosis ofARDS in patients with severe sepsis provides importantnew information about the relative performance of a var-iety of plasma biomarkers of different aspects of the

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

1 Specificity

Sen

sitiv

ity

Figure 1 Receiver operator characteristic (ROC) curve analysisof the plasma biomarker panels for differentiating ARDS (cases)from controls. Predicted probability of ARDS for each subject wascomputed from a logistic regression model that includes the topfive biomarkers (SP-D, RAGE, IL-8, CC-16 and IL-6). Specificity andsensitivity were computed at each possible cutoff of the predictedprobability. Three ROC analyses are shown. The solid line showsthe ROC analysis for all 200 patients in the study (100 cases, 100controls). The AUC is 0.75 (95% CI: 0.7 to 0.84). The dashed lineshows the ROC analysis using only the 91 cases who had ARDS atthe time of the blood draw for biomarker measurement as well astheir matched controls. The AUC for this model is 0.78 (95% CI:0.74 to 0.87). The dotted line shows the ROC analysis using onlythe 66 patients who had the most severe ARDS (ARDS on all studydays) and their matched controls. The AUC for this model is 0.82(95% CI: 0.77 to 0.90). ROC, receiver operator characteristic curve;ARDS, acute respiratory distress syndrome; SP-D, surfactant proteinD; RAGE, receptor for advanced glycation endproducts; IL-8, inter-leukin 8; CC16, club cell protein-16; IL-6, interleukin 6; AUC, areaunder the receiver operator characteristic curve; CI, confidenceinterval.


pathophysiology of ARDS, information that has not previ-ously been available from the many single biomarker stud-ies in ARDS.

ConclusionsIn conclusion, abnormal levels of five biomarkers inplasma provided excellent discrimination for the diagnosisof ARDS in patients with severe sepsis as assessed byROC curve analysis. Three of the five biomarkers weregenerated by the lung epithelium, suggesting that lungepithelial injury is a critical determinant of alveolar flood-ing and the subsequent arterial hypoxemia and bilateralopacities that constitute the clinical definition of ARDS, afinding that is concordant with evidence that impaired al-veolar epithelial fluid clearance is characteristic of patientswith ARDS [31,32]. Although the definition of ARDSis based on clinical criteria, altered levels of plasma bio-markers may be useful to assist in confirming the diagno-sis in patients with shock and possible sepsis, and alsopotentially selecting patients for clinical trials that are de-signed to reduce lung epithelial injury [5,28]. The bio-marker panel might also be useful to categorize patientswith sepsis-induced ARDS by their biologic profile as wellas their clinical profile, as has been done recently in otherlung disease, such as asthma [33].

Key messages

! Among a panel of 11 biomarkers of various aspectsof the pathophysiology of ARDS, biomarkers of lungepithelial injury and inflammation were the mostuseful for discriminating sepsis patients with ARDSfrom those without ARDS.

! A five biomarker panel that included SP-D, RAGE,CC-16, IL-8 and IL-6 had an area under the ROC

curve of 0.75 (95% CI: 0.7 to 0.84) for diagnosis ofARDS.

! For diagnosis of more severe ARDS the area underthe ROC curve was 0.82 (95% CI: 0.77 to 0.90).

! Altered levels of plasma biomarkers may be usefulbiologic confirmation of the diagnosis of ARDS inpatients with sepsis.

! A biomarker panel that includes biomarkers of lungepithelial injury and inflammation may be useful forselecting patients for clinical trials that are designedto reduce lung epithelial injury.

AbbreviationsAECC: American European consensus conference; ALI: Acute lung injury;ANG-2: Angiopoietin-2; APACHE II: Acute physiology and chronic healthevaluation II; ARDS: Acute respiratory distress syndrome; AUC: Area under thereceiver operator characteristic curve; BNP: Brain natriuretic peptide;CC16: Club cell secretory protein; ELISA: Enzyme- linked immunosorbentassay; IL-6: Interleukin 6; IL-8: Interleukin 8; MMP: Matrix metalloprotease;PCPIII: Procollagen peptide III; RAGE: Receptor for advanced glycationendproducts; ROC: Receiver operator characteristic; SAPS II: Simplified acutephysiology score II; SP-D: Surfactant protein D; TNF-α: Tumor necrosis factoralpha; VALID: Validation of acute lung injury biomarkers for diagnosis study.

Competing interestsThe authors declare that they have no competing interests.

Authors’ contributionsLBW designed the study, oversaw the data acquisition and analysis and draftedthe manuscript. TK and ZZ designed the study and performed the data analysis.DR collected and interpreted patient data for the study. NW made the biomarkermeasurements. GB designed the study and assisted with interpretation of results.AM enrolled patients in the study and assisted with interpretation of results. CCand MM designed the study and assisted with interpretation of results. Allauthors read and approved the final manuscript.

AcknowledgementsThis study was supported by funding to LBW from the National Institutes ofHealth (NIH HL081332, HL103836) and an American Heart Association EstablishedInvestigator Award. The funding bodies had no role in the study design, in thecollection, analysis, and interpretation of data; in the writing of the manuscript orin the decision to submit the manuscript for publication.

Points0 10 20 30 40 50 60 70 80 90 100

SPD8 16 34 57 117 304 572

RAGE300 500 900 2800 7800

IL82 5 23 1084 9600

CC16146 69 22 11 6 3 1

IL67 16 77 220 707 9033 39719

Total Points0 20 40 60 80 100 120 140 160 180 200 220 240

Probability of ARDS0.1 0.2 0.4 0.6 0.8 0.9

Figure 2 The multivariable logistic regression model was used to create a prediction model nomogram for the probability of ARDS. A valuein each biomarker predictor variable corresponds to a point scale at the top. The sum of the individual predictor variable points corresponds to the totalpoints and the probability of ARDS shown at the bottom. For each predictor variable, the shown values are approximately 1st, 5th, 25th, 50th, 75th, 95th,and 99th percentiles. ARDS, acute respiratory distress syndrome.


RESEARCH Open Access

Biomarkers of lung epithelial injury andinflammation distinguish severe sepsis patientswith acute respiratory distress syndromeLorraine B Ware1,2*, Tatsuki Koyama3, Zhiguo Zhao3, David R Janz1, Nancy Wickersham1, Gordon R Bernard1,Addison K May4, Carolyn S Calfee5 and Michael A Matthay5

Abstract

Introduction: Despite recent modifications, the clinical definition of the acute respiratory distress syndrome (ARDS)remains non-specific, leading to under-diagnosis and under-treatment. This study was designed to test the hypothesisthat a biomarker panel would be useful for biologic confirmation of the clinical diagnosis of ARDS in patients at risk ofdeveloping ARDS due to severe sepsis.

Methods: This was a retrospective case control study of 100 patients with severe sepsis and no evidence of ARDScompared to 100 patients with severe sepsis and evidence of ARDS on at least two of their first four ICU days. A panelthat included 11 biomarkers of inflammation, fibroblast activation, proteolytic injury, endothelial injury, and lungepithelial injury was measured in plasma from the morning of ICU day two. A backward elimination model buildingstrategy on 1,000 bootstrapped data was used to select the best performing biomarkers for further consideration in alogistic regression model for diagnosis of ARDS.

Results: Using the five best-performing biomarkers (surfactant protein-D (SP-D), receptor for advanced glycationend-products (RAGE), interleukin-8 (IL-8), club cell secretory protein (CC-16), and interleukin-6 (IL-6)) the area underthe receiver operator characteristic curve (AUC) was 0.75 (95% CI: 0.7 to 0.84) for the diagnosis of ARDS. The AUCimproved to 0.82 (95% CI: 0.77 to 0.90) for diagnosis of severe ARDS, defined as ARDS present on all four of thefirst four ICU days.

Conclusions: Abnormal levels of five plasma biomarkers including three biomarkers generated by lung epithelium(SP-D, RAGE, CC-16) provided excellent discrimination for diagnosis of ARDS in patients with severe sepsis. Alteredlevels of plasma biomarkers may be useful biologic confirmation of the diagnosis of ARDS in patients with sepsis,and also potentially for selecting patients for clinical trials that are designed to reduce lung epithelial injury.

IntroductionThe Acute Respiratory Distress Syndrome (ARDS) is acommon clinical syndrome of acute lung inflammation,non-cardiogenic pulmonary edema and acute respira-tory failure [1]. Despite recent modifications [2] to theAmerican European Consensus Conference (AECC)definition [3], the clinical definition of ARDS remains

non-specific and is not uniformly applied. As a result,ARDS remains underdiagnosed and undertreated.The discovery and validation of biomarkers of myocar-

dial injury and ventricular overload such as troponinand brain-natriuretic peptide (BNP) has transformed thediagnosis, management and design of clinical trials inconditions such as myocardial infarction and congestiveheart failure. In a similar way, identification of plasmabiomarkers that facilitate diagnosis of ARDS could improveclinical care, enhance our understanding of pathophysi-ology, and could be used to enroll a more homogeneousgroup of patients into clinical trials of new therapies,increasing the likelihood of detecting a treatment effect.Although several plasma biomarkers have been studied in

* Correspondence: [email protected] of Allergy, Pulmonary and Critical Care Medicine, Department ofMedicine, Vanderbilt University School of Medicine, T1218 MCN, 1161 21stAvenue S, Nashville, TN 37232-2650, USA2Department of Pathology, Microbiology and Immunology, VanderbiltUniversity School of Medicine, Nashville, TN, USAFull list of author information is available at the end of the article

© 2013 Ware et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.

Ware et al. Critical Care 2013, 17:R253http://ccforum.com/content/17/5/R253

Wareetal.Cri>calCare2013,17:R253

Jensen et al. Ann. Intensive Care (2016) 6:114 DOI 10.1186/s13613-016-0212-y

RESEARCH

Prediction of non-recovery from ventilator-demanding acute respiratory failure, ARDS and death using lung damage biomarkers: data from a 1200-patient critical care randomized trialJens-Ulrik S. Jensen1,2* , Theis S. Itenov1,3, Katrin M. Thormar4,5, Lars Hein3,6, Thomas T. Mohr5,6, Mads H. Andersen7, Jesper Løken8, Hamid Tousi9, Bettina Lundgren2, Hans Christian Boesen6, Maria E. Johansen1, Sisse R. Ostrowski10, Pär I. Johansson10, Jesper Grarup1, Jørgen Vestbo11, Jens D. Lundgren1 and For The Procalcitonin And Survival Study (PASS) Group

Abstract Background: It is unclear whether biomarkers of alveolar damage (surfactant protein D, SPD) or conductive airway damage (club cell secretory protein 16, CC16) measured early after intensive care admittance are associated with one-month clinical respiratory prognosis. If patients who do not recover respiratory function within one month can be identified early, future experimental lung interventions can be aimed toward this high-risk group. We aimed to determine, in a heterogenous critically ill population, whether baseline profound alveolar damage or conductive airway damage has clinical respiratory impact one month after intensive care admittance.

Methods: Biobank study of biomarkers of alveolar and conductive airway damage in intensive care patients was conducted. This was a sub-study of 758 intubated patients from a 1200-patient randomized trial. We split the cohort into a “learning cohort” and “validating cohort” based on geographical criteria: northern sites (learning) and southern sites (validating).

Results: Baseline SPD above the 85th percentile in the “learning cohort” predicted low chance of successful wean-ing from ventilator within 28 days (adjusted hazard ratio 0.6 [95% CI 0.4–0.9], p = 0.005); this was confirmed in the validating cohort. CC16 did not predict the endpoint. The absolute risk of not being successfully weaned within the first month was 48/106 (45.3%) vs. 175/652 (26.8%), p < 0.0001 (high SPD vs. low SPD). The chance of being “alive and without ventilator ≥20 days within the first month” was lower among patients with high SPD (adjusted OR 0.2 [95% CI 0.2–0.4], p < 0.0001), confirmed in the validating cohort, and the risk of ARDS was higher among patients with high SPD (adjusted OR 3.4 [95% CI 1.0–11.4], p = 0.04)—also confirmed in the validating cohort.

Conclusion: Early profound alveolar damage in intubated patients can be identified by SPD blood measurement at intensive care admission, and high SPD level is a strong independent predictor that the patient suffers from ARDS and will not recover independent respiratory function within one month. This knowledge can be used to improve

© The Author(s) 2016. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Open Access

*Correspondence: [email protected] 1 CHIP/Department of Infectious Diseases, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, Copenhagen Ø, DenmarkFull list of author information is available at the end of the article

Page 6 of 11Jensen et al. Ann. Intensive Care (2016) 6:114

known or suspected predictors of respiratory failure [22, 23]. Additionally, the association is biologically plausible, since SPD is a documented part of the alveolar surfactant with immunological properties as a pattern recognition

molecule, and serum SPD increases when alveolar dam-age emerges so a leakage to the blood can occur [7, 24, 25]. Almost half of the patients with high SPD were not capable of breathing on own conditions after one month

Fig. 2 Surfactant protein D serum levels according to primary admission reason. a “Learning cohort”/northern sites. b “Validating cohort.” Boxes are medians and interquartile ranges. Whiskers are total range. Rhombuses are means

Table 2 Predictors of successful weaning from mechanical ventilation within 28 days—multivariable competing risk Cox regression

Adjusted Cox regression risk estimates for known, suspected and explored predictors of successful weaning from ventilator within 28 days. Death from all causes was entered in the model as a competing risk

Q1, quartile 1; eGFR, ml/min/1.73 m2

Learning cohort (northern) N = 405 Validating cohort (southern) N = 353

P value Hazard ratio 95% CI for HR P value Hazard ratio 95% CI for HR

Lower Upper Lower Upper

Surfactant protein D (≥85th percentile in “learning cohort,” ≥525.6 ng/mL vs. <525.6 ng/mL)

0.0053 0.60 0.42 0.86 0.046 0.64 0.42 0.99

Club cell secretory protein 16 (≥85th percentile in “learning cohort,” ≥42.8 ng/mL vs. <42.8 ng/mL)

0.50 0.89 0.66 1.22 0.81 0.96 0.67 1.37

PaO2/FiO2 (Q1 vs. Q2–Q4) 0.017 0.73 0.57 0.95 0.89 0.98 0.73 1.32

Apache II score (per score unit increase) 0.031 0.986 0.974 0.999 0.041 0.981 0.964 0.999

Age (per year increase) 0.84 1.00 0.992 1.010 0.17 0.993 0.984 1.003

Severe sepsis/Septic shock (vs. milder or no infection) 0.0014 0.66 0.51 0.85 0.0019 0.63 0.48 0.85

Charlson’s comorbidity index ≥2 vs. <2 0.046 1.29 1.00 1.67 0.58 1.08 0.81 1.46

Chronic obstructive pulmonary disease (yes vs. no) 0.36 0.87 0.64 1.18 0.71 0.94 0.67 1.31

Gender (male vs. female) 0.82 0.97 0.77 1.23 0.97 1.01 0.77 1.32

Estimated glomerular filtration rate (per ml) 0.61 1.00 0.999 1.002 0.048 1.001 0.998 1.004

Jensenetal.Ann.IntensiveCare(2016)6:114

of 11Jensen et al. Ann. Intensive Care (2016) 6:114

was reached, compared to approximately one-fourth of those with SPD below the cutoff.

Opposite, CC16, a marker primarily produced in the conductive airways, did not predict the endpoint in any analysis.

The current results should be interpreted in context to the not easily recognized pathophysiological changes taking place in acute ventilator-dependent respiratory failure, sometimes leading to the clinical picture of acute respiratory distress syndrome (ARDS): diffuse damage to alveolar epithelial cells, including alveolar type II cells and vascular endothelium, breakdown of the basal mem-brane in the alveoli and consequent leakage of surfactant components to the blood. This is accompanied by hem-orrhagic intra-alveolar deposition of platelets, protein and fibrin components eventually forming hyaline mem-branes [26].

Thus, our results indicate that patients who have an early increase in SPD more often progress into patho-physiological changes that are not easily reversible and changes that cause the phenotype of persistent ventila-tor-dependent respiratory failure. Our findings regarding PaO2/FiO2 ratio, an important acute parameter for the

ICU physician, underline the need for pathophysiologi-cal markers like SPD to identify patients at early risk of persistent respiratory failure, since an unfavorable PaO2/FiO2 ratio in the lowest quartile was not a consistent pre-dictor of poor respiratory prognosis after one month.

The knowledge provided by measuring SPD early does have important implications for predicting outcome, and it does increase the understanding of how and when the decisive pathophysiological steps leading to this feared clinical syndrome occur, but even more importantly, this knowledge could help at admission, identifying the most suitable candidates for trials applying experimen-tal lung interventions in patients at high risk of develop-ing persistent respiratory failure. Prone positioning has been demonstrated to be effective in patients with ARDS [27]; however, our results suggest that patients with a predicted high risk of persistent respiratory failure (i.e., highest SPD) should be enrolled in trials testing experi-mental lung interventions even before ARDS develops, in order to improve the prognosis [28]. ARDS awareness is of key importance in these vulnerable patients, so timely and effective interventions can be initiated. However, current reports show that ARDS is often not recognized, even when present [22]. Additionally, many patients who end up with persistent respiratory failure after one month may not have fulfilled ARDS criteria previous to this, and in some patients, an early warning by a bio-marker, before ARDS develops, may provide a possibility for early intervention, even before clinical signs of poor prognosis can be realized. Thus, it seems reasonable to supplement increasing ARDS awareness with biomarkers of acute lung damage like SPD and probably others. In a rat model of ARDS, soluble receptor for advanced gly-cation end products (sRAGE) seemed to reflect alveolar type I cell injury, and this was also observed in humans [29, 30], and recently, it has been demonstrated that a strong negative correlation exists between alveolar fluid clearance rate and plasma sRAGE in a murine model as well as in humans [31]. Thus, sRAGE and SPD may pro-vide complementary information on the pathophysiologi-cal changes taking place in the alveolar epithelium.

This knowledge does draw the attention to two issues of pivotal importance: i) that novel and experimental alveoli-protecting interventions should be instituted in a personalized manner—what works for one critically ill patient (with altered SPD, sRAGE and possibly other sig-nals of profound lung damage) may not work for another patient with low SPD and no other significant signs of profound lung damage, and ii) that in patients with early signs of profound lung damage, experimental alveoli-pro-tecting interventions should probably be tested in trials to reduce development of long-term respiratory failure.

Fig. 3 Cumulative incidence of successful weaning from respira-tor within 28 days after intensive care admission and death while intubated—total cohort (“learning”/northern cohort + “validating”/southern cohort). The two upper curves are regarding “successful weaning from respirator” (vs. still intubated at day 28); the two lower curves are regarding “dead while intubated” (vs. alive at day 28). Patients extubated <48 h at death were counted as “dead while intu-bated.” Patients extubated and alive at day 28 and those extubated ≥48 h at death were counted as successfully weaned from ventilator. N = 758. Gray scales are 95% CI. SPD surfactant protein D. “High SPD is >525.6 ng/mL

Jensenetal.Ann.IntensiveCare(2016)6:114

et al studied 192 trauma patients with and without ALI andwho all had blood drawn during their acute illness to deter-mine if a panel of biomarkers would improve the diagnosticaccuracy in differentiating ALI from cardiogenic pulmonaryedema. Twenty-one biomarkers were measured, seven of

which were ultimately included in a diagnostic model dueto their individual discriminating characteristics in differen-tiating ALI from cardiogenic pulmonary edema. These sevenbiomarkers (brain natriuretic peptide, RAGE, procollagenpeptide III, Ang-2, IL-8, IL-10, and TNF-α) performed verywell in differentiating ALI from control patients, with an AUCof 0.86 (►Fig. 1).76Combining biomarkers into panels to aid inrisk assessment, diagnosis, and prognosis is an emergingfocus in the study of ALI and has the potential to cross overnot only to bedside use but also to increase the accuracy ofpatient selection in future clinical trials throughmore preciseidentification of patients with ALI/ARDS, identification ofthose who have a poor prognosis, and those who may bemore likely to benefit from new therapies.77,78

Future Discovery of Biomarkers

The biomarker studies that have been discussed here arehypothesis driven and focused on biomarkers identified fromour current understanding of the pathogenesis of ALI/ARDS.This targeted approach ignores potential mediators of lunginjury that have yet to be recognized as important in thepathogenesis of ALI/ARDS. Unbiased methodologies, includ-ing metabolomics, proteomics, gene expression analysis, andgenome-wide association studies have the potential to iden-tify new biomarkers of ALI/ARDS as well as new mediatorsand pathways that are mechanistically important (►Table 2).

MetabolomicsMetabolomics is the study of the products (metabolites) ofspecific cellular processes. Profiles of these downstreamproducts can be created to describe cellular, tissue, or organdisease processes. Metabolomics differs from proteomics andgenomics in that metabolomics aims to study the end

Fig. 1 Receiver operating characteristic (ROC) curve analysis of theseven best biomarkers differentiating trauma-induced acute lunginjury/acute respiratory distress syndrome (ALI/ARDS) from controls,along with a panel of three biomarkers (receptor for advancedglycation end-product [RAGE], brain natriuretic peptide [BNP], pro-collagen peptide III [PCP III]) with similar discriminatory character-istics. (From Fremont RD, Koyama T, Calfee CS, et al. Acute lung injuryin patients with traumatic injuries: utility of a panel of biomarkers fordiagnosis and pathogenesis. J Trauma 2010;68:1124; withpermission.)

Table 2 Novel methodologies for biomarker discovery

Method Strengths Weaknesses

Metabolomics Unbiased biomarker discovery Normal ranges not described

Metabolites are more easily measurable at thebedside

Metabolite levels very low in edema fluid

Describes downstream function of proteinalterations

Metabolites may be affected by processesother than disease

Proteomics Discovery of proteins not previously considered inacute lung injury

Extensive lists of proteins without descriptionof pathogenesis

Proteins are downstream effectors of geneticvariability

Sensitivity for low abundance proteins is stillproblematic

Identified proteins may serve as clinical biomarkersfor diagnosis or prognosis

High abundance proteins may overwhelmdisease-related signals

Gene expression profiling Changes in gene expression may provide moreinsight into pathogenesis than genetic variation

Difficult to obtain lung tissue so most studieshave been done in whole blood and primarilyreflect changes in leukocyte gene expression

Genome-wideassociation studies

Identification of genetic underpinnings of disease Functional significance of genetic variabilitymay be difficult to ascertain

Illuminates previously undescribed pathways Adjustment for multiple comparisonsrequired extremely large sample sizes

Seminars in Respiratory and Critical Care Medicine Vol. 34 No. 4/2013

Biomarkers of ALI/ARDS Janz, Ware542

This

doc

umen

t was

dow

nloa

ded

for p

erso

nal u

se o

nly.

Una

utho

rized

dis

tribu

tion

is s

trict

ly p

rohi

bite

d.

SeminRespirCritCareMed2013;34:537–548

PhysiolGenomics37:133–139,2009.

CONCLUSIONES •  Los biomarcadores en lesiones pulmonares agudas han proporcionado valiososconocimientossobrelapatogénesis.

•  Enlosúl>mosdiezaños,sehanprobadovariosestudiosdebiomarcadoresengrandesestudios.

•  Un único biomarcador o paneles de marcadores que están fácilmente disponibles yprediceneldesarrollodeALIodiagnos>canALIparausoclínicoderu>nasiguensiendodikcilesdealcanzar.

•  Conlamejoradelasplataformasdealtorendimientoyladisponibilidaddeherramientasde bioinformá>ca cada vezmás sofis>cadas, existe una gran esperanza de iden>ficarnuevas firmas de genes y proteínas o moléculas pequeñas que sirvan comobiomarcadoresparalapredicción,elpronós>coyeldiagnós>codeALI.

•  Es de esperar que estos hallazgos proporcionen información sobre la biología de laenfermedadeiden>fiquenobje>vosnuevosparalasintervencionesterapéu>cas.

sÍndrome de insuficiencia respiratoria aguda:...

Documents